diff options
Diffstat (limited to 'servers/rendering/renderer_rd')
68 files changed, 6468 insertions, 3261 deletions
diff --git a/servers/rendering/renderer_rd/cluster_builder_rd.h b/servers/rendering/renderer_rd/cluster_builder_rd.h index ebb81abdad..c0c03eb26a 100644 --- a/servers/rendering/renderer_rd/cluster_builder_rd.h +++ b/servers/rendering/renderer_rd/cluster_builder_rd.h @@ -235,7 +235,7 @@ public: Transform3D xform = view_xform * p_transform; float radius = xform.basis.get_uniform_scale(); - if (radius > 0.98 || radius < 1.02) { + if (radius < 0.98 || radius > 1.02) { xform.basis.orthonormalize(); } diff --git a/servers/rendering/renderer_rd/effects_rd.cpp b/servers/rendering/renderer_rd/effects_rd.cpp index 111c75debd..236eb5e596 100644 --- a/servers/rendering/renderer_rd/effects_rd.cpp +++ b/servers/rendering/renderer_rd/effects_rd.cpp @@ -37,19 +37,8 @@ #include "servers/rendering/renderer_rd/renderer_compositor_rd.h" #include "thirdparty/misc/cubemap_coeffs.h" -static _FORCE_INLINE_ void store_transform_3x3(const Basis &p_basis, float *p_array) { - p_array[0] = p_basis.elements[0][0]; - p_array[1] = p_basis.elements[1][0]; - p_array[2] = p_basis.elements[2][0]; - p_array[3] = 0; - p_array[4] = p_basis.elements[0][1]; - p_array[5] = p_basis.elements[1][1]; - p_array[6] = p_basis.elements[2][1]; - p_array[7] = 0; - p_array[8] = p_basis.elements[0][2]; - p_array[9] = p_basis.elements[1][2]; - p_array[10] = p_basis.elements[2][2]; - p_array[11] = 0; +bool EffectsRD::get_prefer_raster_effects() { + return prefer_raster_effects; } static _FORCE_INLINE_ void store_camera(const CameraMatrix &p_mtx, float *p_array) { @@ -81,6 +70,28 @@ RID EffectsRD::_get_uniform_set_from_image(RID p_image) { return uniform_set; } +RID EffectsRD::_get_uniform_set_for_input(RID p_texture) { + if (input_to_uniform_set_cache.has(p_texture)) { + RID uniform_set = input_to_uniform_set_cache[p_texture]; + if (RD::get_singleton()->uniform_set_is_valid(uniform_set)) { + return uniform_set; + } + } + + Vector<RD::Uniform> uniforms; + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_INPUT_ATTACHMENT; + u.binding = 0; + u.ids.push_back(p_texture); + uniforms.push_back(u); + // This is specific to our subpass shader + RID uniform_set = RD::get_singleton()->uniform_set_create(uniforms, tonemap.shader.version_get_shader(tonemap.shader_version, TONEMAP_MODE_SUBPASS), 0); + + input_to_uniform_set_cache[p_texture] = uniform_set; + + return uniform_set; +} + RID EffectsRD::_get_uniform_set_from_texture(RID p_texture, bool p_use_mipmaps) { if (texture_to_uniform_set_cache.has(p_texture)) { RID uniform_set = texture_to_uniform_set_cache[p_texture]; @@ -96,7 +107,7 @@ RID EffectsRD::_get_uniform_set_from_texture(RID p_texture, bool p_use_mipmaps) u.ids.push_back(p_use_mipmaps ? default_mipmap_sampler : default_sampler); u.ids.push_back(p_texture); uniforms.push_back(u); - //anything with the same configuration (one texture in binding 0 for set 0), is good + // anything with the same configuration (one texture in binding 0 for set 0), is good RID uniform_set = RD::get_singleton()->uniform_set_create(uniforms, tonemap.shader.version_get_shader(tonemap.shader_version, 0), 0); texture_to_uniform_set_cache[p_texture] = uniform_set; @@ -398,6 +409,8 @@ void EffectsRD::set_color(RID p_dest_texture, const Color &p_color, const Rect2i } void EffectsRD::gaussian_blur(RID p_source_rd_texture, RID p_texture, RID p_back_texture, const Rect2i &p_region, bool p_8bit_dst) { + ERR_FAIL_COND_MSG(!prefer_raster_effects, "Can't use the compute version of the gaussian blur with the mobile renderer."); + memset(©.push_constant, 0, sizeof(CopyPushConstant)); uint32_t base_flags = 0; @@ -431,6 +444,8 @@ void EffectsRD::gaussian_blur(RID p_source_rd_texture, RID p_texture, RID p_back } void EffectsRD::gaussian_glow(RID p_source_rd_texture, RID p_back_texture, const Size2i &p_size, float p_strength, bool p_high_quality, bool p_first_pass, float p_luminance_cap, float p_exposure, float p_bloom, float p_hdr_bleed_treshold, float p_hdr_bleed_scale, RID p_auto_exposure, float p_auto_exposure_grey) { + ERR_FAIL_COND_MSG(prefer_raster_effects, "Can't use the compute version of the gaussian glow with the mobile renderer."); + memset(©.push_constant, 0, sizeof(CopyPushConstant)); CopyMode copy_mode = p_first_pass && p_auto_exposure.is_valid() ? COPY_MODE_GAUSSIAN_GLOW_AUTO_EXPOSURE : COPY_MODE_GAUSSIAN_GLOW; @@ -464,6 +479,57 @@ void EffectsRD::gaussian_glow(RID p_source_rd_texture, RID p_back_texture, const RD::get_singleton()->compute_list_end(); } +void EffectsRD::gaussian_glow_raster(RID p_source_rd_texture, RID p_framebuffer_half, RID p_rd_texture_half, RID p_dest_framebuffer, const Vector2 &p_pixel_size, float p_strength, bool p_high_quality, bool p_first_pass, float p_luminance_cap, float p_exposure, float p_bloom, float p_hdr_bleed_treshold, float p_hdr_bleed_scale, RID p_auto_exposure, float p_auto_exposure_grey) { + ERR_FAIL_COND_MSG(!prefer_raster_effects, "Can't use the raster version of the gaussian glow with the clustered renderer."); + + memset(&blur_raster.push_constant, 0, sizeof(BlurRasterPushConstant)); + + BlurRasterMode blur_mode = p_first_pass && p_auto_exposure.is_valid() ? BLUR_MODE_GAUSSIAN_GLOW_AUTO_EXPOSURE : BLUR_MODE_GAUSSIAN_GLOW; + uint32_t base_flags = 0; + + blur_raster.push_constant.pixel_size[0] = p_pixel_size.x; + blur_raster.push_constant.pixel_size[1] = p_pixel_size.y; + + blur_raster.push_constant.glow_strength = p_strength; + blur_raster.push_constant.glow_bloom = p_bloom; + blur_raster.push_constant.glow_hdr_threshold = p_hdr_bleed_treshold; + blur_raster.push_constant.glow_hdr_scale = p_hdr_bleed_scale; + blur_raster.push_constant.glow_exposure = p_exposure; + blur_raster.push_constant.glow_white = 0; //actually unused + blur_raster.push_constant.glow_luminance_cap = p_luminance_cap; + + blur_raster.push_constant.glow_auto_exposure_grey = p_auto_exposure_grey; //unused also + + //HORIZONTAL + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_framebuffer_half, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, blur_raster.pipelines[blur_mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_framebuffer_half))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_source_rd_texture), 0); + if (p_auto_exposure.is_valid() && p_first_pass) { + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_auto_exposure), 1); + } + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); + + blur_raster.push_constant.flags = base_flags | BLUR_FLAG_HORIZONTAL | (p_first_pass ? BLUR_FLAG_GLOW_FIRST_PASS : 0); + RD::get_singleton()->draw_list_set_push_constant(draw_list, &blur_raster.push_constant, sizeof(BlurRasterPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); + + blur_mode = BLUR_MODE_GAUSSIAN_GLOW; + + //VERTICAL + draw_list = RD::get_singleton()->draw_list_begin(p_dest_framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, blur_raster.pipelines[blur_mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_dest_framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_rd_texture_half), 0); + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); + + blur_raster.push_constant.flags = base_flags; + RD::get_singleton()->draw_list_set_push_constant(draw_list, &blur_raster.push_constant, sizeof(BlurRasterPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); +} + void EffectsRD::screen_space_reflection(RID p_diffuse, RID p_normal_roughness, RenderingServer::EnvironmentSSRRoughnessQuality p_roughness_quality, RID p_blur_radius, RID p_blur_radius2, RID p_metallic, const Color &p_metallic_mask, RID p_depth, RID p_scale_depth, RID p_scale_normal, RID p_output, RID p_output_blur, const Size2i &p_screen_size, int p_max_steps, float p_fade_in, float p_fade_out, float p_tolerance, const CameraMatrix &p_camera) { RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); @@ -518,11 +584,10 @@ void EffectsRD::screen_space_reflection(RID p_diffuse, RID p_normal_roughness, R if (p_roughness_quality != RS::ENV_SSR_ROUGNESS_QUALITY_DISABLED) { RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_image_pair(p_output, p_blur_radius), 1); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture_pair(p_metallic, p_normal_roughness), 3); } else { RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_output), 1); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_metallic), 3); } + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_metallic), 3); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_scale_normal), 2); RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.width, p_screen_size.height, 1); @@ -674,7 +739,27 @@ void EffectsRD::make_mipmap(RID p_source_rd_texture, RID p_dest_texture, const S RD::get_singleton()->compute_list_end(); } -void EffectsRD::copy_cubemap_to_dp(RID p_source_rd_texture, RID p_dst_framebuffer, const Rect2 &p_rect, float p_z_near, float p_z_far, bool p_dp_flip) { +void EffectsRD::make_mipmap_raster(RID p_source_rd_texture, RID p_dest_framebuffer, const Size2i &p_size) { + ERR_FAIL_COND_MSG(!prefer_raster_effects, "Can't use the raster version of mipmap with the clustered renderer."); + + memset(&blur_raster.push_constant, 0, sizeof(BlurRasterPushConstant)); + + BlurRasterMode mode = BLUR_MIPMAP; + + blur_raster.push_constant.pixel_size[0] = 1.0 / float(p_size.x); + blur_raster.push_constant.pixel_size[1] = 1.0 / float(p_size.y); + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_dest_framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, blur_raster.pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_dest_framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_source_rd_texture), 0); + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); + RD::get_singleton()->draw_list_set_push_constant(draw_list, &blur_raster.push_constant, sizeof(BlurRasterPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); +} + +void EffectsRD::copy_cubemap_to_dp(RID p_source_rd_texture, RID p_dst_framebuffer, const Rect2 &p_rect, const Vector2 &p_dst_size, float p_z_near, float p_z_far, bool p_dp_flip) { CopyToDPPushConstant push_constant; push_constant.screen_rect[0] = p_rect.position.x; push_constant.screen_rect[1] = p_rect.position.y; @@ -682,7 +767,9 @@ void EffectsRD::copy_cubemap_to_dp(RID p_source_rd_texture, RID p_dst_framebuffe push_constant.screen_rect[3] = p_rect.size.height; push_constant.z_far = p_z_far; push_constant.z_near = p_z_near; - push_constant.z_flip = p_dp_flip; + push_constant.texel_size[0] = 1.0f / p_dst_size.x; + push_constant.texel_size[1] = 1.0f / p_dst_size.y; + push_constant.texel_size[0] *= p_dp_flip ? -1.0f : 1.0f; // Encode dp flip as x size sign RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_dst_framebuffer, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ); RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, cube_to_dp.pipeline.get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_dst_framebuffer))); @@ -725,6 +812,7 @@ void EffectsRD::tonemapper(RID p_source_color, RID p_dst_framebuffer, const Tone tonemap.push_constant.exposure = p_settings.exposure; tonemap.push_constant.white = p_settings.white; tonemap.push_constant.auto_exposure_grey = p_settings.auto_exposure_grey; + tonemap.push_constant.luminance_multiplier = p_settings.luminance_multiplier; tonemap.push_constant.use_color_correction = p_settings.use_color_correction; @@ -735,11 +823,11 @@ void EffectsRD::tonemapper(RID p_source_color, RID p_dst_framebuffer, const Tone if (p_settings.view_count > 1) { // Use MULTIVIEW versions - mode += 4; + mode += 6; } RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_dst_framebuffer, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD); - RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, tonemap.pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_dst_framebuffer))); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, tonemap.pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_dst_framebuffer), false, RD::get_singleton()->draw_list_get_current_pass())); RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_source_color), 0); RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_settings.exposure_texture), 1); RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_settings.glow_texture, true), 2); @@ -751,7 +839,49 @@ void EffectsRD::tonemapper(RID p_source_color, RID p_dst_framebuffer, const Tone RD::get_singleton()->draw_list_end(); } +void EffectsRD::tonemapper(RD::DrawListID p_subpass_draw_list, RID p_source_color, RD::FramebufferFormatID p_dst_format_id, const TonemapSettings &p_settings) { + memset(&tonemap.push_constant, 0, sizeof(TonemapPushConstant)); + + tonemap.push_constant.use_bcs = p_settings.use_bcs; + tonemap.push_constant.bcs[0] = p_settings.brightness; + tonemap.push_constant.bcs[1] = p_settings.contrast; + tonemap.push_constant.bcs[2] = p_settings.saturation; + + ERR_FAIL_COND_MSG(p_settings.use_glow, "Glow is not supported when using subpasses."); + tonemap.push_constant.use_glow = p_settings.use_glow; + + int mode = p_settings.use_1d_color_correction ? TONEMAP_MODE_SUBPASS_1D_LUT : TONEMAP_MODE_SUBPASS; + if (p_settings.view_count > 1) { + // Use MULTIVIEW versions + mode += 6; + } + + tonemap.push_constant.tonemapper = p_settings.tonemap_mode; + tonemap.push_constant.use_auto_exposure = p_settings.use_auto_exposure; + tonemap.push_constant.exposure = p_settings.exposure; + tonemap.push_constant.white = p_settings.white; + tonemap.push_constant.auto_exposure_grey = p_settings.auto_exposure_grey; + + tonemap.push_constant.use_color_correction = p_settings.use_color_correction; + + tonemap.push_constant.use_debanding = p_settings.use_debanding; + tonemap.push_constant.luminance_multiplier = p_settings.luminance_multiplier; + + RD::get_singleton()->draw_list_bind_render_pipeline(p_subpass_draw_list, tonemap.pipelines[mode].get_render_pipeline(RD::INVALID_ID, p_dst_format_id, false, RD::get_singleton()->draw_list_get_current_pass())); + RD::get_singleton()->draw_list_bind_uniform_set(p_subpass_draw_list, _get_uniform_set_for_input(p_source_color), 0); + RD::get_singleton()->draw_list_bind_uniform_set(p_subpass_draw_list, _get_uniform_set_from_texture(p_settings.exposure_texture), 1); // should be set to a default texture, it's ignored + RD::get_singleton()->draw_list_bind_uniform_set(p_subpass_draw_list, _get_uniform_set_from_texture(p_settings.glow_texture, true), 2); // should be set to a default texture, it's ignored + RD::get_singleton()->draw_list_bind_uniform_set(p_subpass_draw_list, _get_uniform_set_from_texture(p_settings.color_correction_texture), 3); + + RD::get_singleton()->draw_list_bind_index_array(p_subpass_draw_list, index_array); + + RD::get_singleton()->draw_list_set_push_constant(p_subpass_draw_list, &tonemap.push_constant, sizeof(TonemapPushConstant)); + RD::get_singleton()->draw_list_draw(p_subpass_draw_list, true); +} + void EffectsRD::luminance_reduction(RID p_source_texture, const Size2i p_source_size, const Vector<RID> p_reduce, RID p_prev_luminance, float p_min_luminance, float p_max_luminance, float p_adjust, bool p_set) { + ERR_FAIL_COND_MSG(prefer_raster_effects, "Can't use compute version of luminance reduction with the mobile renderer."); + luminance_reduce.push_constant.source_size[0] = p_source_size.x; luminance_reduce.push_constant.source_size[1] = p_source_size.y; luminance_reduce.push_constant.max_luminance = p_max_luminance; @@ -790,7 +920,41 @@ void EffectsRD::luminance_reduction(RID p_source_texture, const Size2i p_source_ RD::get_singleton()->compute_list_end(); } -void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i &p_base_texture_size, RID p_secondary_texture, RID p_halfsize_texture1, RID p_halfsize_texture2, bool p_dof_far, float p_dof_far_begin, float p_dof_far_size, bool p_dof_near, float p_dof_near_begin, float p_dof_near_size, float p_bokeh_size, RenderingServer::DOFBokehShape p_bokeh_shape, RS::DOFBlurQuality p_quality, bool p_use_jitter, float p_cam_znear, float p_cam_zfar, bool p_cam_orthogonal) { +void EffectsRD::luminance_reduction_raster(RID p_source_texture, const Size2i p_source_size, const Vector<RID> p_reduce, Vector<RID> p_fb, RID p_prev_luminance, float p_min_luminance, float p_max_luminance, float p_adjust, bool p_set) { + ERR_FAIL_COND_MSG(!prefer_raster_effects, "Can't use raster version of luminance reduction with the clustered renderer."); + ERR_FAIL_COND_MSG(p_reduce.size() != p_fb.size(), "Incorrect frame buffer account for luminance reduction."); + + luminance_reduce_raster.push_constant.max_luminance = p_max_luminance; + luminance_reduce_raster.push_constant.min_luminance = p_min_luminance; + luminance_reduce_raster.push_constant.exposure_adjust = p_adjust; + + for (int i = 0; i < p_reduce.size(); i++) { + luminance_reduce_raster.push_constant.source_size[0] = i == 0 ? p_source_size.x : luminance_reduce_raster.push_constant.dest_size[0]; + luminance_reduce_raster.push_constant.source_size[1] = i == 0 ? p_source_size.y : luminance_reduce_raster.push_constant.dest_size[1]; + luminance_reduce_raster.push_constant.dest_size[0] = MAX(luminance_reduce_raster.push_constant.source_size[0] / 8, 1); + luminance_reduce_raster.push_constant.dest_size[1] = MAX(luminance_reduce_raster.push_constant.source_size[1] / 8, 1); + + bool final = !p_set && (luminance_reduce_raster.push_constant.dest_size[0] == 1) && (luminance_reduce_raster.push_constant.dest_size[1] == 1); + LuminanceReduceRasterMode mode = final ? LUMINANCE_REDUCE_FRAGMENT_FINAL : (i == 0 ? LUMINANCE_REDUCE_FRAGMENT_FIRST : LUMINANCE_REDUCE_FRAGMENT); + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_fb[i], RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, luminance_reduce_raster.pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_fb[i]))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(i == 0 ? p_source_texture : p_reduce[i - 1]), 0); + if (final) { + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_prev_luminance), 1); + } + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &luminance_reduce_raster.push_constant, sizeof(LuminanceReduceRasterPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); + } +} + +void EffectsRD::bokeh_dof(const BokehBuffers &p_buffers, bool p_dof_far, float p_dof_far_begin, float p_dof_far_size, bool p_dof_near, float p_dof_near_begin, float p_dof_near_size, float p_bokeh_size, RenderingServer::DOFBokehShape p_bokeh_shape, RS::DOFBlurQuality p_quality, bool p_use_jitter, float p_cam_znear, float p_cam_zfar, bool p_cam_orthogonal) { + ERR_FAIL_COND_MSG(prefer_raster_effects, "Can't use compute version of BOKEH DOF with the mobile renderer."); + bokeh.push_constant.blur_far_active = p_dof_far; bokeh.push_constant.blur_far_begin = p_dof_far_begin; bokeh.push_constant.blur_far_end = p_dof_far_begin + p_dof_far_size; @@ -817,22 +981,22 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i // The alpha channel of the source color texture is filled with the expected circle size // If used for DOF far, the size is positive, if used for near, its negative. - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, bokeh.pipelines[BOKEH_GEN_BLUR_SIZE]); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, bokeh.compute_pipelines[BOKEH_GEN_BLUR_SIZE]); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_base_texture), 0); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_depth_texture), 1); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_buffers.base_texture), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_buffers.depth_texture), 1); - bokeh.push_constant.size[0] = p_base_texture_size.x; - bokeh.push_constant.size[1] = p_base_texture_size.y; + bokeh.push_constant.size[0] = p_buffers.base_texture_size.x; + bokeh.push_constant.size[1] = p_buffers.base_texture_size.y; RD::get_singleton()->compute_list_set_push_constant(compute_list, &bokeh.push_constant, sizeof(BokehPushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_base_texture_size.x, p_base_texture_size.y, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_buffers.base_texture_size.x, p_buffers.base_texture_size.y, 1); RD::get_singleton()->compute_list_add_barrier(compute_list); if (p_bokeh_shape == RS::DOF_BOKEH_BOX || p_bokeh_shape == RS::DOF_BOKEH_HEXAGON) { //second pass - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, bokeh.pipelines[p_bokeh_shape == RS::DOF_BOKEH_BOX ? BOKEH_GEN_BOKEH_BOX : BOKEH_GEN_BOKEH_HEXAGONAL]); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, bokeh.compute_pipelines[p_bokeh_shape == RS::DOF_BOKEH_BOX ? BOKEH_GEN_BOKEH_BOX : BOKEH_GEN_BOKEH_HEXAGONAL]); static const int quality_samples[4] = { 6, 12, 12, 24 }; @@ -841,18 +1005,18 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i if (p_quality == RS::DOF_BLUR_QUALITY_VERY_LOW || p_quality == RS::DOF_BLUR_QUALITY_LOW) { //box and hexagon are more or less the same, and they can work in either half (very low and low quality) or full (medium and high quality_ sizes) - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_halfsize_texture1), 0); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_base_texture), 1); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_buffers.half_texture[0]), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_buffers.base_texture), 1); - bokeh.push_constant.size[0] = p_base_texture_size.x >> 1; - bokeh.push_constant.size[1] = p_base_texture_size.y >> 1; + bokeh.push_constant.size[0] = p_buffers.base_texture_size.x >> 1; + bokeh.push_constant.size[1] = p_buffers.base_texture_size.y >> 1; bokeh.push_constant.half_size = true; bokeh.push_constant.blur_size *= 0.5; } else { //medium and high quality use full size - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_secondary_texture), 0); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_base_texture), 1); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_buffers.secondary_texture), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_buffers.base_texture), 1); } RD::get_singleton()->compute_list_set_push_constant(compute_list, &bokeh.push_constant, sizeof(BokehPushConstant)); @@ -864,11 +1028,11 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i bokeh.push_constant.second_pass = true; if (p_quality == RS::DOF_BLUR_QUALITY_VERY_LOW || p_quality == RS::DOF_BLUR_QUALITY_LOW) { - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_halfsize_texture2), 0); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_halfsize_texture1), 1); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_buffers.half_texture[1]), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_buffers.half_texture[0]), 1); } else { - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_base_texture), 0); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_secondary_texture), 1); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_buffers.base_texture), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_buffers.secondary_texture), 1); } RD::get_singleton()->compute_list_set_push_constant(compute_list, &bokeh.push_constant, sizeof(BokehPushConstant)); @@ -879,25 +1043,25 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i if (p_quality == RS::DOF_BLUR_QUALITY_VERY_LOW || p_quality == RS::DOF_BLUR_QUALITY_LOW) { //forth pass, upscale for low quality - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, bokeh.pipelines[BOKEH_COMPOSITE]); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, bokeh.compute_pipelines[BOKEH_COMPOSITE]); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_base_texture), 0); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_halfsize_texture2), 1); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_buffers.base_texture), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_buffers.half_texture[1]), 1); - bokeh.push_constant.size[0] = p_base_texture_size.x; - bokeh.push_constant.size[1] = p_base_texture_size.y; + bokeh.push_constant.size[0] = p_buffers.base_texture_size.x; + bokeh.push_constant.size[1] = p_buffers.base_texture_size.y; bokeh.push_constant.half_size = false; bokeh.push_constant.second_pass = false; RD::get_singleton()->compute_list_set_push_constant(compute_list, &bokeh.push_constant, sizeof(BokehPushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_base_texture_size.x, p_base_texture_size.y, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_buffers.base_texture_size.x, p_buffers.base_texture_size.y, 1); } } else { //circle //second pass - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, bokeh.pipelines[BOKEH_GEN_BOKEH_CIRCULAR]); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, bokeh.compute_pipelines[BOKEH_GEN_BOKEH_CIRCULAR]); static const float quality_scale[4] = { 8.0, 4.0, 1.0, 0.5 }; @@ -906,11 +1070,11 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i //circle always runs in half size, otherwise too expensive - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_halfsize_texture1), 0); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_base_texture), 1); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_buffers.half_texture[0]), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_buffers.base_texture), 1); - bokeh.push_constant.size[0] = p_base_texture_size.x >> 1; - bokeh.push_constant.size[1] = p_base_texture_size.y >> 1; + bokeh.push_constant.size[0] = p_buffers.base_texture_size.x >> 1; + bokeh.push_constant.size[1] = p_buffers.base_texture_size.y >> 1; bokeh.push_constant.half_size = true; RD::get_singleton()->compute_list_set_push_constant(compute_list, &bokeh.push_constant, sizeof(BokehPushConstant)); @@ -922,28 +1086,202 @@ void EffectsRD::bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i // upscale - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, bokeh.pipelines[BOKEH_COMPOSITE]); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, bokeh.compute_pipelines[BOKEH_COMPOSITE]); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_base_texture), 0); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_halfsize_texture1), 1); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_buffers.base_texture), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_buffers.half_texture[0]), 1); - bokeh.push_constant.size[0] = p_base_texture_size.x; - bokeh.push_constant.size[1] = p_base_texture_size.y; + bokeh.push_constant.size[0] = p_buffers.base_texture_size.x; + bokeh.push_constant.size[1] = p_buffers.base_texture_size.y; bokeh.push_constant.half_size = false; bokeh.push_constant.second_pass = false; RD::get_singleton()->compute_list_set_push_constant(compute_list, &bokeh.push_constant, sizeof(BokehPushConstant)); - RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_base_texture_size.x, p_base_texture_size.y, 1); + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_buffers.base_texture_size.x, p_buffers.base_texture_size.y, 1); } RD::get_singleton()->compute_list_end(); } -void EffectsRD::gather_ssao(RD::ComputeListID p_compute_list, const Vector<RID> p_ao_slices, const SSAOSettings &p_settings, bool p_adaptive_base_pass) { - RD::get_singleton()->compute_list_bind_uniform_set(p_compute_list, ssao.gather_uniform_set, 0); +void EffectsRD::bokeh_dof_raster(const BokehBuffers &p_buffers, bool p_dof_far, float p_dof_far_begin, float p_dof_far_size, bool p_dof_near, float p_dof_near_begin, float p_dof_near_size, float p_dof_blur_amount, RenderingServer::DOFBokehShape p_bokeh_shape, RS::DOFBlurQuality p_quality, float p_cam_znear, float p_cam_zfar, bool p_cam_orthogonal) { + ERR_FAIL_COND_MSG(!prefer_raster_effects, "Can't use blur DOF with the clustered renderer."); + + memset(&bokeh.push_constant, 0, sizeof(BokehPushConstant)); + + bokeh.push_constant.orthogonal = p_cam_orthogonal; + bokeh.push_constant.size[0] = p_buffers.base_texture_size.width; + bokeh.push_constant.size[1] = p_buffers.base_texture_size.height; + bokeh.push_constant.z_far = p_cam_zfar; + bokeh.push_constant.z_near = p_cam_znear; + + bokeh.push_constant.second_pass = false; + bokeh.push_constant.half_size = false; + bokeh.push_constant.blur_size = p_dof_blur_amount; + + if (p_dof_far || p_dof_near) { + if (p_dof_far) { + bokeh.push_constant.blur_far_active = true; + bokeh.push_constant.blur_far_begin = p_dof_far_begin; + bokeh.push_constant.blur_far_end = p_dof_far_begin + p_dof_far_size; + } + + if (p_dof_near) { + bokeh.push_constant.blur_near_active = true; + bokeh.push_constant.blur_near_begin = p_dof_near_begin; + bokeh.push_constant.blur_near_end = p_dof_near_begin - p_dof_near_size; + } + + { + // generate our depth data + RID framebuffer = p_buffers.base_weight_fb; + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, bokeh.raster_pipelines[BOKEH_GEN_BLUR_SIZE].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_buffers.depth_texture), 0); + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &bokeh.push_constant, sizeof(BokehPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); + } + + if (p_bokeh_shape == RS::DOF_BOKEH_BOX || p_bokeh_shape == RS::DOF_BOKEH_HEXAGON) { + // double pass approach + BokehMode mode = p_bokeh_shape == RS::DOF_BOKEH_BOX ? BOKEH_GEN_BOKEH_BOX : BOKEH_GEN_BOKEH_HEXAGONAL; + + if (p_quality == RS::DOF_BLUR_QUALITY_VERY_LOW || p_quality == RS::DOF_BLUR_QUALITY_LOW) { + //box and hexagon are more or less the same, and they can work in either half (very low and low quality) or full (medium and high quality_ sizes) + bokeh.push_constant.size[0] = p_buffers.base_texture_size.x >> 1; + bokeh.push_constant.size[1] = p_buffers.base_texture_size.y >> 1; + bokeh.push_constant.half_size = true; + bokeh.push_constant.blur_size *= 0.5; + } + + static const int quality_samples[4] = { 6, 12, 12, 24 }; + bokeh.push_constant.blur_scale = 0.5; + bokeh.push_constant.steps = quality_samples[p_quality]; + + RID framebuffer = bokeh.push_constant.half_size ? p_buffers.half_fb[0] : p_buffers.secondary_fb; + + // Pass 1 + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, bokeh.raster_pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_buffers.base_texture), 0); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_buffers.weight_texture[0]), 1); + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &bokeh.push_constant, sizeof(BokehPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); + + // Pass 2 + if (!bokeh.push_constant.half_size) { + // do not output weight, we're writing back into our base buffer + mode = p_bokeh_shape == RS::DOF_BOKEH_BOX ? BOKEH_GEN_BOKEH_BOX_NOWEIGHT : BOKEH_GEN_BOKEH_HEXAGONAL_NOWEIGHT; + } + bokeh.push_constant.second_pass = true; + + framebuffer = bokeh.push_constant.half_size ? p_buffers.half_fb[1] : p_buffers.base_fb; + RID texture = bokeh.push_constant.half_size ? p_buffers.half_texture[0] : p_buffers.secondary_texture; + RID weight = bokeh.push_constant.half_size ? p_buffers.weight_texture[2] : p_buffers.weight_texture[1]; + + draw_list = RD::get_singleton()->draw_list_begin(framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, bokeh.raster_pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(texture), 0); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(weight), 1); + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &bokeh.push_constant, sizeof(BokehPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); + + if (bokeh.push_constant.half_size) { + // Compose pass + mode = BOKEH_COMPOSITE; + framebuffer = p_buffers.base_fb; + + draw_list = RD::get_singleton()->draw_list_begin(framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, bokeh.raster_pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_buffers.half_texture[1]), 0); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_buffers.weight_texture[3]), 1); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_buffers.weight_texture[0]), 2); + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &bokeh.push_constant, sizeof(BokehPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); + } + + } else { + // circular is a single pass approach + BokehMode mode = BOKEH_GEN_BOKEH_CIRCULAR; + + { + // circle always runs in half size, otherwise too expensive (though the code below does support making this optional) + bokeh.push_constant.size[0] = p_buffers.base_texture_size.x >> 1; + bokeh.push_constant.size[1] = p_buffers.base_texture_size.y >> 1; + bokeh.push_constant.half_size = true; + // bokeh.push_constant.blur_size *= 0.5; + } + + static const float quality_scale[4] = { 8.0, 4.0, 1.0, 0.5 }; + bokeh.push_constant.blur_scale = quality_scale[p_quality]; + bokeh.push_constant.steps = 0.0; + + RID framebuffer = bokeh.push_constant.half_size ? p_buffers.half_fb[0] : p_buffers.secondary_fb; + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, bokeh.raster_pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_buffers.base_texture), 0); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_buffers.weight_texture[0]), 1); + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &bokeh.push_constant, sizeof(BokehPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); + + if (bokeh.push_constant.half_size) { + // Compose + mode = BOKEH_COMPOSITE; + framebuffer = p_buffers.base_fb; + + draw_list = RD::get_singleton()->draw_list_begin(framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, bokeh.raster_pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_buffers.half_texture[0]), 0); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_buffers.weight_texture[2]), 1); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_buffers.weight_texture[0]), 2); + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &bokeh.push_constant, sizeof(BokehPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); + } else { + // Just copy it back (we use our blur raster shader here).. + draw_list = RD::get_singleton()->draw_list_begin(p_buffers.base_fb, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, blur_raster.pipelines[BLUR_MODE_COPY].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_buffers.base_fb))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_buffers.secondary_texture), 0); + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); + + memset(&blur_raster.push_constant, 0, sizeof(BlurRasterPushConstant)); + RD::get_singleton()->draw_list_set_push_constant(draw_list, &blur_raster.push_constant, sizeof(BlurRasterPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); + } + } + } +} + +void EffectsRD::gather_ssao(RD::ComputeListID p_compute_list, const Vector<RID> p_ao_slices, const SSAOSettings &p_settings, bool p_adaptive_base_pass, RID p_gather_uniform_set, RID p_importance_map_uniform_set) { + RD::get_singleton()->compute_list_bind_uniform_set(p_compute_list, p_gather_uniform_set, 0); if ((p_settings.quality == RS::ENV_SSAO_QUALITY_ULTRA) && !p_adaptive_base_pass) { - RD::get_singleton()->compute_list_bind_uniform_set(p_compute_list, ssao.importance_map_uniform_set, 1); + RD::get_singleton()->compute_list_bind_uniform_set(p_compute_list, p_importance_map_uniform_set, 1); } for (int i = 0; i < 4; i++) { @@ -966,7 +1304,7 @@ void EffectsRD::gather_ssao(RD::ComputeListID p_compute_list, const Vector<RID> RD::get_singleton()->compute_list_add_barrier(p_compute_list); } -void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_depth_mipmaps_texture, const Vector<RID> &p_depth_mipmaps, RID p_ao, const Vector<RID> p_ao_slices, RID p_ao_pong, const Vector<RID> p_ao_pong_slices, RID p_upscale_buffer, RID p_importance_map, RID p_importance_map_pong, const CameraMatrix &p_projection, const SSAOSettings &p_settings, bool p_invalidate_uniform_sets) { +void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_depth_mipmaps_texture, const Vector<RID> &p_depth_mipmaps, RID p_ao, const Vector<RID> p_ao_slices, RID p_ao_pong, const Vector<RID> p_ao_pong_slices, RID p_upscale_buffer, RID p_importance_map, RID p_importance_map_pong, const CameraMatrix &p_projection, const SSAOSettings &p_settings, bool p_invalidate_uniform_sets, RID &r_downsample_uniform_set, RID &r_gather_uniform_set, RID &r_importance_map_uniform_set) { RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); RD::get_singleton()->draw_command_begin_label("SSAO"); /* FIRST PASS */ @@ -996,7 +1334,7 @@ void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_dep u.ids.push_back(p_depth_mipmaps[3]); uniforms.push_back(u); } - ssao.downsample_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, ssao.downsample_shader.version_get_shader(ssao.downsample_shader_version, 2), 2); + r_downsample_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, ssao.downsample_shader.version_get_shader(ssao.downsample_shader_version, 2), 2); } float depth_linearize_mul = -p_projection.matrix[3][2]; @@ -1031,7 +1369,7 @@ void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_dep RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_depth_buffer), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_depth_mipmaps[0]), 1); if (p_settings.quality > RS::ENV_SSAO_QUALITY_MEDIUM) { - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, ssao.downsample_uniform_set, 2); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, r_downsample_uniform_set, 2); } RD::get_singleton()->compute_list_set_push_constant(compute_list, &ssao.downsample_push_constant, sizeof(SSAODownsamplePushConstant)); @@ -1115,7 +1453,7 @@ void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_dep u.ids.push_back(ssao.gather_constants_buffer); uniforms.push_back(u); } - ssao.gather_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, ssao.gather_shader.version_get_shader(ssao.gather_shader_version, 0), 0); + r_gather_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, ssao.gather_shader.version_get_shader(ssao.gather_shader_version, 0), 0); } if (p_invalidate_uniform_sets) { @@ -1142,7 +1480,7 @@ void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_dep u.ids.push_back(ssao.importance_map_load_counter); uniforms.push_back(u); } - ssao.importance_map_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, ssao.gather_shader.version_get_shader(ssao.gather_shader_version, 2), 1); + r_importance_map_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, ssao.gather_shader.version_get_shader(ssao.gather_shader_version, 2), 1); } if (p_settings.quality == RS::ENV_SSAO_QUALITY_ULTRA) { @@ -1153,7 +1491,7 @@ void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_dep ssao.importance_map_push_constant.power = p_settings.power; //base pass RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_GATHER_BASE]); - gather_ssao(compute_list, p_ao_pong_slices, p_settings, true); + gather_ssao(compute_list, p_ao_pong_slices, p_settings, true, r_gather_uniform_set, RID()); //generate importance map RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_GENERATE_IMPORTANCE_MAP]); @@ -1184,7 +1522,7 @@ void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_dep RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, ssao.pipelines[SSAO_GATHER]); } - gather_ssao(compute_list, p_ao_slices, p_settings, false); + gather_ssao(compute_list, p_ao_slices, p_settings, false, r_gather_uniform_set, r_importance_map_uniform_set); RD::get_singleton()->draw_command_end_label(); // Gather SSAO } @@ -1204,8 +1542,9 @@ void EffectsRD::generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_dep if (p_settings.quality > RS::ENV_SSAO_QUALITY_VERY_LOW) { if (pass < blur_passes - 2) { blur_pipeline = SSAO_BLUR_PASS_WIDE; + } else { + blur_pipeline = SSAO_BLUR_PASS_SMART; } - blur_pipeline = SSAO_BLUR_PASS_SMART; } for (int i = 0; i < 4; i++) { @@ -1299,7 +1638,9 @@ void EffectsRD::roughness_limit(RID p_source_normal, RID p_roughness, const Size RD::get_singleton()->compute_list_end(); } -void EffectsRD::cubemap_roughness(RID p_source_rd_texture, RID p_dest_framebuffer, uint32_t p_face_id, uint32_t p_sample_count, float p_roughness, float p_size) { +void EffectsRD::cubemap_roughness(RID p_source_rd_texture, RID p_dest_texture, uint32_t p_face_id, uint32_t p_sample_count, float p_roughness, float p_size) { + ERR_FAIL_COND_MSG(prefer_raster_effects, "Can't use compute based cubemap roughness with the mobile renderer."); + memset(&roughness.push_constant, 0, sizeof(CubemapRoughnessPushConstant)); roughness.push_constant.face_id = p_face_id > 9 ? 0 : p_face_id; @@ -1309,10 +1650,10 @@ void EffectsRD::cubemap_roughness(RID p_source_rd_texture, RID p_dest_framebuffe roughness.push_constant.face_size = p_size; RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, roughness.pipeline); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, roughness.compute_pipeline); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_source_rd_texture), 0); - RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_dest_framebuffer), 1); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_dest_texture), 1); RD::get_singleton()->compute_list_set_push_constant(compute_list, &roughness.push_constant, sizeof(CubemapRoughnessPushConstant)); @@ -1324,11 +1665,37 @@ void EffectsRD::cubemap_roughness(RID p_source_rd_texture, RID p_dest_framebuffe RD::get_singleton()->compute_list_end(); } +void EffectsRD::cubemap_roughness_raster(RID p_source_rd_texture, RID p_dest_framebuffer, uint32_t p_face_id, uint32_t p_sample_count, float p_roughness, float p_size) { + ERR_FAIL_COND_MSG(!prefer_raster_effects, "Can't use raster based cubemap roughness with the clustered renderer."); + ERR_FAIL_COND_MSG(p_face_id >= 6, "Raster implementation of cubemap roughness must process one side at a time."); + + memset(&roughness.push_constant, 0, sizeof(CubemapRoughnessPushConstant)); + + roughness.push_constant.face_id = p_face_id; + roughness.push_constant.roughness = p_roughness; + roughness.push_constant.sample_count = p_sample_count; + roughness.push_constant.use_direct_write = p_roughness == 0.0; + roughness.push_constant.face_size = p_size; + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_dest_framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, roughness.raster_pipeline.get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_dest_framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_source_rd_texture), 0); + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &roughness.push_constant, sizeof(CubemapRoughnessPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); +} + void EffectsRD::cubemap_downsample(RID p_source_cubemap, RID p_dest_cubemap, const Size2i &p_size) { + ERR_FAIL_COND_MSG(prefer_raster_effects, "Can't use compute based cubemap downsample with the mobile renderer."); + cubemap_downsampler.push_constant.face_size = p_size.x; + cubemap_downsampler.push_constant.face_id = 0; // we render all 6 sides to each layer in one call RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, cubemap_downsampler.pipeline); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, cubemap_downsampler.compute_pipeline); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_source_cubemap), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_dest_cubemap), 1); @@ -1342,7 +1709,27 @@ void EffectsRD::cubemap_downsample(RID p_source_cubemap, RID p_dest_cubemap, con RD::get_singleton()->compute_list_end(); } +void EffectsRD::cubemap_downsample_raster(RID p_source_cubemap, RID p_dest_framebuffer, uint32_t p_face_id, const Size2i &p_size) { + ERR_FAIL_COND_MSG(!prefer_raster_effects, "Can't use raster based cubemap downsample with the clustered renderer."); + ERR_FAIL_COND_MSG(p_face_id >= 6, "Raster implementation of cubemap downsample must process one side at a time."); + + cubemap_downsampler.push_constant.face_size = p_size.x; + cubemap_downsampler.push_constant.face_id = p_face_id; + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_dest_framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, cubemap_downsampler.raster_pipeline.get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_dest_framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_source_cubemap), 0); + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &cubemap_downsampler.push_constant, sizeof(CubemapDownsamplerPushConstant)); + + RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); +} + void EffectsRD::cubemap_filter(RID p_source_cubemap, Vector<RID> p_dest_cubemap, bool p_use_array) { + ERR_FAIL_COND_MSG(prefer_raster_effects, "Can't use compute based cubemap filter with the mobile renderer."); + Vector<RD::Uniform> uniforms; for (int i = 0; i < p_dest_cubemap.size(); i++) { RD::Uniform u; @@ -1354,12 +1741,12 @@ void EffectsRD::cubemap_filter(RID p_source_cubemap, Vector<RID> p_dest_cubemap, if (RD::get_singleton()->uniform_set_is_valid(filter.image_uniform_set)) { RD::get_singleton()->free(filter.image_uniform_set); } - filter.image_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, filter.shader.version_get_shader(filter.shader_version, 0), 2); + filter.image_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, filter.compute_shader.version_get_shader(filter.shader_version, 0), 2); int pipeline = p_use_array ? FILTER_MODE_HIGH_QUALITY_ARRAY : FILTER_MODE_HIGH_QUALITY; pipeline = filter.use_high_quality ? pipeline : pipeline + 1; RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); - RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, filter.pipelines[pipeline]); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, filter.compute_pipelines[pipeline]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_source_cubemap, true), 0); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, filter.uniform_set, 1); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, filter.image_uniform_set, 2); @@ -1371,43 +1758,27 @@ void EffectsRD::cubemap_filter(RID p_source_cubemap, Vector<RID> p_dest_cubemap, RD::get_singleton()->compute_list_end(); } -void EffectsRD::render_sky(RD::DrawListID p_list, float p_time, RID p_fb, RID p_samplers, RID p_fog, PipelineCacheRD *p_pipeline, RID p_uniform_set, RID p_texture_set, uint32_t p_view_count, const CameraMatrix *p_projections, const Basis &p_orientation, float p_multiplier, const Vector3 &p_position) { - SkyPushConstant sky_push_constant; - - memset(&sky_push_constant, 0, sizeof(SkyPushConstant)); - - for (uint32_t v = 0; v < p_view_count; v++) { - // We only need key components of our projection matrix - sky_push_constant.projections[v][0] = p_projections[v].matrix[2][0]; - sky_push_constant.projections[v][1] = p_projections[v].matrix[0][0]; - sky_push_constant.projections[v][2] = p_projections[v].matrix[2][1]; - sky_push_constant.projections[v][3] = p_projections[v].matrix[1][1]; - } - sky_push_constant.position[0] = p_position.x; - sky_push_constant.position[1] = p_position.y; - sky_push_constant.position[2] = p_position.z; - sky_push_constant.multiplier = p_multiplier; - sky_push_constant.time = p_time; - store_transform_3x3(p_orientation, sky_push_constant.orientation); - - RenderingDevice::FramebufferFormatID fb_format = RD::get_singleton()->framebuffer_get_format(p_fb); - - RD::DrawListID draw_list = p_list; +void EffectsRD::cubemap_filter_raster(RID p_source_cubemap, RID p_dest_framebuffer, uint32_t p_face_id, uint32_t p_mip_level) { + ERR_FAIL_COND_MSG(!prefer_raster_effects, "Can't use raster based cubemap filter with the clustered renderer."); + ERR_FAIL_COND_MSG(p_face_id >= 6, "Raster implementation of cubemap filter must process one side at a time."); - RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, p_pipeline->get_render_pipeline(RD::INVALID_ID, fb_format)); + // TODO implement! + CubemapFilterRasterPushConstant push_constant; + push_constant.mip_level = p_mip_level; + push_constant.face_id = p_face_id; - RD::get_singleton()->draw_list_bind_uniform_set(draw_list, p_samplers, 0); - if (p_uniform_set.is_valid()) { //material may not have uniform set - RD::get_singleton()->draw_list_bind_uniform_set(draw_list, p_uniform_set, 1); - } - RD::get_singleton()->draw_list_bind_uniform_set(draw_list, p_texture_set, 2); - RD::get_singleton()->draw_list_bind_uniform_set(draw_list, p_fog, 3); + CubemapFilterMode mode = filter.use_high_quality ? FILTER_MODE_HIGH_QUALITY : FILTER_MODE_LOW_QUALITY; + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_dest_framebuffer, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, filter.raster_pipelines[mode].get_render_pipeline(RD::INVALID_ID, RD::get_singleton()->framebuffer_get_format(p_dest_framebuffer))); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, _get_uniform_set_from_texture(p_source_cubemap), 0); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, filter.uniform_set, 1); RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); - RD::get_singleton()->draw_list_set_push_constant(draw_list, &sky_push_constant, sizeof(SkyPushConstant)); + RD::get_singleton()->draw_list_set_push_constant(draw_list, &push_constant, sizeof(CubemapFilterRasterPushConstant)); RD::get_singleton()->draw_list_draw(draw_list, true); + RD::get_singleton()->draw_list_end(); } void EffectsRD::resolve_gi(RID p_source_depth, RID p_source_normal_roughness, RID p_source_voxel_gi, RID p_dest_depth, RID p_dest_normal_roughness, RID p_dest_voxel_gi, Vector2i p_screen_size, int p_samples, uint32_t p_barrier) { @@ -1432,6 +1803,24 @@ void EffectsRD::resolve_gi(RID p_source_depth, RID p_source_normal_roughness, RI RD::get_singleton()->compute_list_end(p_barrier); } +void EffectsRD::resolve_depth(RID p_source_depth, RID p_dest_depth, Vector2i p_screen_size, int p_samples, uint32_t p_barrier) { + ResolvePushConstant push_constant; + push_constant.screen_size[0] = p_screen_size.x; + push_constant.screen_size[1] = p_screen_size.y; + push_constant.samples = p_samples; + + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(); + RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, resolve.pipelines[RESOLVE_MODE_DEPTH]); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_compute_uniform_set_from_texture(p_source_depth), 0); + RD::get_singleton()->compute_list_bind_uniform_set(compute_list, _get_uniform_set_from_image(p_dest_depth), 1); + + RD::get_singleton()->compute_list_set_push_constant(compute_list, &push_constant, sizeof(ResolvePushConstant)); + + RD::get_singleton()->compute_list_dispatch_threads(compute_list, p_screen_size.x, p_screen_size.y, 1); + + RD::get_singleton()->compute_list_end(p_barrier); +} + void EffectsRD::sort_buffer(RID p_uniform_set, int p_size) { Sort::PushConstant push_constant; push_constant.total_elements = p_size; @@ -1501,8 +1890,35 @@ void EffectsRD::sort_buffer(RID p_uniform_set, int p_size) { RD::get_singleton()->compute_list_end(); } -EffectsRD::EffectsRD() { - { // Initialize copy +EffectsRD::EffectsRD(bool p_prefer_raster_effects) { + prefer_raster_effects = p_prefer_raster_effects; + + if (prefer_raster_effects) { + // init blur shader (on compute use copy shader) + + Vector<String> blur_modes; + blur_modes.push_back("\n#define MODE_MIPMAP\n"); // BLUR_MIPMAP + blur_modes.push_back("\n#define MODE_GAUSSIAN_BLUR\n"); // BLUR_MODE_GAUSSIAN_BLUR + blur_modes.push_back("\n#define MODE_GAUSSIAN_GLOW\n"); // BLUR_MODE_GAUSSIAN_GLOW + blur_modes.push_back("\n#define MODE_GAUSSIAN_GLOW\n#define GLOW_USE_AUTO_EXPOSURE\n"); // BLUR_MODE_GAUSSIAN_GLOW_AUTO_EXPOSURE + blur_modes.push_back("\n#define MODE_COPY\n"); // BLUR_MODE_COPY + + blur_raster.shader.initialize(blur_modes); + memset(&blur_raster.push_constant, 0, sizeof(BlurRasterPushConstant)); + blur_raster.shader_version = blur_raster.shader.version_create(); + + for (int i = 0; i < BLUR_MODE_MAX; i++) { + blur_raster.pipelines[i].setup(blur_raster.shader.version_get_shader(blur_raster.shader_version, i), RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), RD::PipelineColorBlendState::create_disabled(), 0); + } + + } else { + // not used in clustered + for (int i = 0; i < BLUR_MODE_MAX; i++) { + blur_raster.pipelines[i].clear(); + } + } + + if (!prefer_raster_effects) { // Initialize copy Vector<String> copy_modes; copy_modes.push_back("\n#define MODE_GAUSSIAN_BLUR\n"); copy_modes.push_back("\n#define MODE_GAUSSIAN_BLUR\n#define DST_IMAGE_8BIT\n"); @@ -1520,10 +1936,21 @@ EffectsRD::EffectsRD() { copy.shader.initialize(copy_modes); memset(©.push_constant, 0, sizeof(CopyPushConstant)); + + if (prefer_raster_effects) { + // disable shaders we can't use + copy.shader.set_variant_enabled(COPY_MODE_GAUSSIAN_COPY, false); + copy.shader.set_variant_enabled(COPY_MODE_GAUSSIAN_COPY_8BIT, false); + copy.shader.set_variant_enabled(COPY_MODE_GAUSSIAN_GLOW, false); + copy.shader.set_variant_enabled(COPY_MODE_GAUSSIAN_GLOW_AUTO_EXPOSURE, false); + } + copy.shader_version = copy.shader.version_create(); for (int i = 0; i < COPY_MODE_MAX; i++) { - copy.pipelines[i] = RD::get_singleton()->compute_pipeline_create(copy.shader.version_get_shader(copy.shader_version, i)); + if (copy.shader.is_variant_enabled(i)) { + copy.pipelines[i] = RD::get_singleton()->compute_pipeline_create(copy.shader.version_get_shader(copy.shader_version, i)); + } } } { @@ -1547,11 +1974,22 @@ EffectsRD::EffectsRD() { // Initialize roughness Vector<String> cubemap_roughness_modes; cubemap_roughness_modes.push_back(""); - roughness.shader.initialize(cubemap_roughness_modes); - roughness.shader_version = roughness.shader.version_create(); + if (prefer_raster_effects) { + roughness.raster_shader.initialize(cubemap_roughness_modes); + + roughness.shader_version = roughness.raster_shader.version_create(); - roughness.pipeline = RD::get_singleton()->compute_pipeline_create(roughness.shader.version_get_shader(roughness.shader_version, 0)); + roughness.raster_pipeline.setup(roughness.raster_shader.version_get_shader(roughness.shader_version, 0), RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), RD::PipelineColorBlendState::create_disabled(), 0); + + } else { + roughness.compute_shader.initialize(cubemap_roughness_modes); + + roughness.shader_version = roughness.compute_shader.version_create(); + + roughness.compute_pipeline = RD::get_singleton()->compute_pipeline_create(roughness.compute_shader.version_get_shader(roughness.shader_version, 0)); + roughness.raster_pipeline.clear(); + } } { @@ -1561,12 +1999,16 @@ EffectsRD::EffectsRD() { tonemap_modes.push_back("\n#define USE_GLOW_FILTER_BICUBIC\n"); tonemap_modes.push_back("\n#define USE_1D_LUT\n"); tonemap_modes.push_back("\n#define USE_GLOW_FILTER_BICUBIC\n#define USE_1D_LUT\n"); + tonemap_modes.push_back("\n#define SUBPASS\n"); + tonemap_modes.push_back("\n#define SUBPASS\n#define USE_1D_LUT\n"); // multiview versions of our shaders tonemap_modes.push_back("\n#define MULTIVIEW\n"); tonemap_modes.push_back("\n#define MULTIVIEW\n#define USE_GLOW_FILTER_BICUBIC\n"); tonemap_modes.push_back("\n#define MULTIVIEW\n#define USE_1D_LUT\n"); tonemap_modes.push_back("\n#define MULTIVIEW\n#define USE_GLOW_FILTER_BICUBIC\n#define USE_1D_LUT\n"); + tonemap_modes.push_back("\n#define MULTIVIEW\n#define SUBPASS\n"); + tonemap_modes.push_back("\n#define MULTIVIEW\n#define SUBPASS\n#define USE_1D_LUT\n"); tonemap.shader.initialize(tonemap_modes); @@ -1575,6 +2017,8 @@ EffectsRD::EffectsRD() { tonemap.shader.set_variant_enabled(TONEMAP_MODE_BICUBIC_GLOW_FILTER_MULTIVIEW, false); tonemap.shader.set_variant_enabled(TONEMAP_MODE_1D_LUT_MULTIVIEW, false); tonemap.shader.set_variant_enabled(TONEMAP_MODE_BICUBIC_GLOW_FILTER_1D_LUT_MULTIVIEW, false); + tonemap.shader.set_variant_enabled(TONEMAP_MODE_SUBPASS_MULTIVIEW, false); + tonemap.shader.set_variant_enabled(TONEMAP_MODE_SUBPASS_1D_LUT_MULTIVIEW, false); } tonemap.shader_version = tonemap.shader.version_create(); @@ -1588,7 +2032,20 @@ EffectsRD::EffectsRD() { } } - { + if (prefer_raster_effects) { + Vector<String> luminance_reduce_modes; + luminance_reduce_modes.push_back("\n#define FIRST_PASS\n"); // LUMINANCE_REDUCE_FRAGMENT_FIRST + luminance_reduce_modes.push_back("\n"); // LUMINANCE_REDUCE_FRAGMENT + luminance_reduce_modes.push_back("\n#define FINAL_PASS\n"); // LUMINANCE_REDUCE_FRAGMENT_FINAL + + luminance_reduce_raster.shader.initialize(luminance_reduce_modes); + memset(&luminance_reduce_raster.push_constant, 0, sizeof(LuminanceReduceRasterPushConstant)); + luminance_reduce_raster.shader_version = luminance_reduce_raster.shader.version_create(); + + for (int i = 0; i < LUMINANCE_REDUCE_FRAGMENT_MAX; i++) { + luminance_reduce_raster.pipelines[i].setup(luminance_reduce_raster.shader.version_get_shader(luminance_reduce_raster.shader_version, i), RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), RD::PipelineColorBlendState::create_disabled(), 0); + } + } else { // Initialize luminance_reduce Vector<String> luminance_reduce_modes; luminance_reduce_modes.push_back("\n#define READ_TEXTURE\n"); @@ -1602,6 +2059,10 @@ EffectsRD::EffectsRD() { for (int i = 0; i < LUMINANCE_REDUCE_MAX; i++) { luminance_reduce.pipelines[i] = RD::get_singleton()->compute_pipeline_create(luminance_reduce.shader.version_get_shader(luminance_reduce.shader_version, i)); } + + for (int i = 0; i < LUMINANCE_REDUCE_FRAGMENT_MAX; i++) { + luminance_reduce_raster.pipelines[i].clear(); + } } { @@ -1620,25 +2081,43 @@ EffectsRD::EffectsRD() { cube_to_dp.pipeline.setup(shader, RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), dss, RD::PipelineColorBlendState(), 0); } - { - // Initialize bokeh - Vector<String> bokeh_modes; - bokeh_modes.push_back("\n#define MODE_GEN_BLUR_SIZE\n"); - bokeh_modes.push_back("\n#define MODE_BOKEH_BOX\n"); - bokeh_modes.push_back("\n#define MODE_BOKEH_HEXAGONAL\n"); - bokeh_modes.push_back("\n#define MODE_BOKEH_CIRCULAR\n"); - bokeh_modes.push_back("\n#define MODE_COMPOSITE_BOKEH\n"); + // Initialize bokeh + Vector<String> bokeh_modes; + bokeh_modes.push_back("\n#define MODE_GEN_BLUR_SIZE\n"); + bokeh_modes.push_back("\n#define MODE_BOKEH_BOX\n#define OUTPUT_WEIGHT\n"); + bokeh_modes.push_back("\n#define MODE_BOKEH_BOX\n"); + bokeh_modes.push_back("\n#define MODE_BOKEH_HEXAGONAL\n#define OUTPUT_WEIGHT\n"); + bokeh_modes.push_back("\n#define MODE_BOKEH_HEXAGONAL\n"); + bokeh_modes.push_back("\n#define MODE_BOKEH_CIRCULAR\n#define OUTPUT_WEIGHT\n"); + bokeh_modes.push_back("\n#define MODE_COMPOSITE_BOKEH\n"); + if (prefer_raster_effects) { + bokeh.raster_shader.initialize(bokeh_modes); - bokeh.shader.initialize(bokeh_modes); + bokeh.shader_version = bokeh.raster_shader.version_create(); - bokeh.shader_version = bokeh.shader.version_create(); + const int att_count[BOKEH_MAX] = { 1, 2, 1, 2, 1, 2, 1 }; + for (int i = 0; i < BOKEH_MAX; i++) { + RD::PipelineColorBlendState blend_state = (i == BOKEH_COMPOSITE) ? RD::PipelineColorBlendState::create_blend(att_count[i]) : RD::PipelineColorBlendState::create_disabled(att_count[i]); + bokeh.raster_pipelines[i].setup(bokeh.raster_shader.version_get_shader(bokeh.shader_version, i), RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), blend_state, 0); + } + } else { + bokeh.compute_shader.initialize(bokeh_modes); + bokeh.compute_shader.set_variant_enabled(BOKEH_GEN_BOKEH_BOX_NOWEIGHT, false); + bokeh.compute_shader.set_variant_enabled(BOKEH_GEN_BOKEH_HEXAGONAL_NOWEIGHT, false); + bokeh.shader_version = bokeh.compute_shader.version_create(); + + for (int i = 0; i < BOKEH_MAX; i++) { + if (bokeh.compute_shader.is_variant_enabled(i)) { + bokeh.compute_pipelines[i] = RD::get_singleton()->compute_pipeline_create(bokeh.compute_shader.version_get_shader(bokeh.shader_version, i)); + } + } for (int i = 0; i < BOKEH_MAX; i++) { - bokeh.pipelines[i] = RD::get_singleton()->compute_pipeline_create(bokeh.shader.version_get_shader(bokeh.shader_version, i)); + bokeh.raster_pipelines[i].clear(); } } - { + if (!prefer_raster_effects) { // Initialize ssao RD::SamplerState sampler; @@ -1694,10 +2173,8 @@ EffectsRD::EffectsRD() { for (int pass = 0; pass < 4; pass++) { for (int subPass = 0; subPass < sub_pass_count; subPass++) { int a = pass; - int b = subPass; - int spmap[5]{ 0, 1, 4, 3, 2 }; - b = spmap[subPass]; + int b = spmap[subPass]; float ca, sa; float angle0 = (float(a) + float(b) / float(sub_pass_count)) * Math_PI * 0.5f; @@ -1782,7 +2259,7 @@ EffectsRD::EffectsRD() { ERR_FAIL_COND(pipeline != SSAO_MAX); } - { + if (!prefer_raster_effects) { // Initialize roughness limiter Vector<String> shader_modes; shader_modes.push_back(""); @@ -1798,11 +2275,21 @@ EffectsRD::EffectsRD() { //Initialize cubemap downsampler Vector<String> cubemap_downsampler_modes; cubemap_downsampler_modes.push_back(""); - cubemap_downsampler.shader.initialize(cubemap_downsampler_modes); - cubemap_downsampler.shader_version = cubemap_downsampler.shader.version_create(); + if (prefer_raster_effects) { + cubemap_downsampler.raster_shader.initialize(cubemap_downsampler_modes); + + cubemap_downsampler.shader_version = cubemap_downsampler.raster_shader.version_create(); - cubemap_downsampler.pipeline = RD::get_singleton()->compute_pipeline_create(cubemap_downsampler.shader.version_get_shader(cubemap_downsampler.shader_version, 0)); + cubemap_downsampler.raster_pipeline.setup(cubemap_downsampler.raster_shader.version_get_shader(cubemap_downsampler.shader_version, 0), RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), RD::PipelineColorBlendState::create_disabled(), 0); + } else { + cubemap_downsampler.compute_shader.initialize(cubemap_downsampler_modes); + + cubemap_downsampler.shader_version = cubemap_downsampler.compute_shader.version_create(); + + cubemap_downsampler.compute_pipeline = RD::get_singleton()->compute_pipeline_create(cubemap_downsampler.compute_shader.version_get_shader(cubemap_downsampler.shader_version, 0)); + cubemap_downsampler.raster_pipeline.clear(); + } } { @@ -1814,12 +2301,6 @@ EffectsRD::EffectsRD() { cubemap_filter_modes.push_back("\n#define USE_LOW_QUALITY\n"); cubemap_filter_modes.push_back("\n#define USE_HIGH_QUALITY\n#define USE_TEXTURE_ARRAY\n"); cubemap_filter_modes.push_back("\n#define USE_LOW_QUALITY\n#define USE_TEXTURE_ARRAY\n"); - filter.shader.initialize(cubemap_filter_modes); - filter.shader_version = filter.shader.version_create(); - - for (int i = 0; i < FILTER_MODE_MAX; i++) { - filter.pipelines[i] = RD::get_singleton()->compute_pipeline_create(filter.shader.version_get_shader(filter.shader_version, i)); - } if (filter.use_high_quality) { filter.coefficient_buffer = RD::get_singleton()->storage_buffer_create(sizeof(high_quality_coeffs)); @@ -1829,18 +2310,54 @@ EffectsRD::EffectsRD() { RD::get_singleton()->buffer_update(filter.coefficient_buffer, 0, sizeof(low_quality_coeffs), &low_quality_coeffs[0]); } - Vector<RD::Uniform> uniforms; - { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; - u.binding = 0; - u.ids.push_back(filter.coefficient_buffer); - uniforms.push_back(u); + if (prefer_raster_effects) { + filter.raster_shader.initialize(cubemap_filter_modes); + + // array variants are not supported in raster + filter.raster_shader.set_variant_enabled(FILTER_MODE_HIGH_QUALITY_ARRAY, false); + filter.raster_shader.set_variant_enabled(FILTER_MODE_LOW_QUALITY_ARRAY, false); + + filter.shader_version = filter.raster_shader.version_create(); + + for (int i = 0; i < FILTER_MODE_MAX; i++) { + if (filter.raster_shader.is_variant_enabled(i)) { + filter.raster_pipelines[i].setup(filter.raster_shader.version_get_shader(filter.shader_version, i), RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), RD::PipelineColorBlendState::create_disabled(), 0); + } else { + filter.raster_pipelines[i].clear(); + } + } + + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 0; + u.ids.push_back(filter.coefficient_buffer); + uniforms.push_back(u); + } + filter.uniform_set = RD::get_singleton()->uniform_set_create(uniforms, filter.raster_shader.version_get_shader(filter.shader_version, filter.use_high_quality ? 0 : 1), 1); + } else { + filter.compute_shader.initialize(cubemap_filter_modes); + filter.shader_version = filter.compute_shader.version_create(); + + for (int i = 0; i < FILTER_MODE_MAX; i++) { + filter.compute_pipelines[i] = RD::get_singleton()->compute_pipeline_create(filter.compute_shader.version_get_shader(filter.shader_version, i)); + filter.raster_pipelines[i].clear(); + } + + Vector<RD::Uniform> uniforms; + { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; + u.binding = 0; + u.ids.push_back(filter.coefficient_buffer); + uniforms.push_back(u); + } + filter.uniform_set = RD::get_singleton()->uniform_set_create(uniforms, filter.compute_shader.version_get_shader(filter.shader_version, filter.use_high_quality ? 0 : 1), 1); } - filter.uniform_set = RD::get_singleton()->uniform_set_create(uniforms, filter.shader.version_get_shader(filter.shader_version, filter.use_high_quality ? 0 : 1), 1); } - { + if (!prefer_raster_effects) { Vector<String> specular_modes; specular_modes.push_back("\n#define MODE_MERGE\n"); specular_modes.push_back("\n#define MODE_MERGE\n#define MODE_SSR\n"); @@ -1876,71 +2393,74 @@ EffectsRD::EffectsRD() { } } - { - Vector<String> ssr_modes; - ssr_modes.push_back("\n"); - ssr_modes.push_back("\n#define MODE_ROUGH\n"); + if (!prefer_raster_effects) { + { + Vector<String> ssr_modes; + ssr_modes.push_back("\n"); + ssr_modes.push_back("\n#define MODE_ROUGH\n"); - ssr.shader.initialize(ssr_modes); + ssr.shader.initialize(ssr_modes); - ssr.shader_version = ssr.shader.version_create(); + ssr.shader_version = ssr.shader.version_create(); - for (int i = 0; i < SCREEN_SPACE_REFLECTION_MAX; i++) { - ssr.pipelines[i] = RD::get_singleton()->compute_pipeline_create(ssr.shader.version_get_shader(ssr.shader_version, i)); + for (int i = 0; i < SCREEN_SPACE_REFLECTION_MAX; i++) { + ssr.pipelines[i] = RD::get_singleton()->compute_pipeline_create(ssr.shader.version_get_shader(ssr.shader_version, i)); + } } - } - { - Vector<String> ssr_filter_modes; - ssr_filter_modes.push_back("\n"); - ssr_filter_modes.push_back("\n#define VERTICAL_PASS\n"); + { + Vector<String> ssr_filter_modes; + ssr_filter_modes.push_back("\n"); + ssr_filter_modes.push_back("\n#define VERTICAL_PASS\n"); - ssr_filter.shader.initialize(ssr_filter_modes); + ssr_filter.shader.initialize(ssr_filter_modes); - ssr_filter.shader_version = ssr_filter.shader.version_create(); + ssr_filter.shader_version = ssr_filter.shader.version_create(); - for (int i = 0; i < SCREEN_SPACE_REFLECTION_FILTER_MAX; i++) { - ssr_filter.pipelines[i] = RD::get_singleton()->compute_pipeline_create(ssr_filter.shader.version_get_shader(ssr_filter.shader_version, i)); + for (int i = 0; i < SCREEN_SPACE_REFLECTION_FILTER_MAX; i++) { + ssr_filter.pipelines[i] = RD::get_singleton()->compute_pipeline_create(ssr_filter.shader.version_get_shader(ssr_filter.shader_version, i)); + } } - } - { - Vector<String> ssr_scale_modes; - ssr_scale_modes.push_back("\n"); + { + Vector<String> ssr_scale_modes; + ssr_scale_modes.push_back("\n"); - ssr_scale.shader.initialize(ssr_scale_modes); + ssr_scale.shader.initialize(ssr_scale_modes); - ssr_scale.shader_version = ssr_scale.shader.version_create(); + ssr_scale.shader_version = ssr_scale.shader.version_create(); - ssr_scale.pipeline = RD::get_singleton()->compute_pipeline_create(ssr_scale.shader.version_get_shader(ssr_scale.shader_version, 0)); - } + ssr_scale.pipeline = RD::get_singleton()->compute_pipeline_create(ssr_scale.shader.version_get_shader(ssr_scale.shader_version, 0)); + } - { - Vector<String> sss_modes; - sss_modes.push_back("\n#define USE_11_SAMPLES\n"); - sss_modes.push_back("\n#define USE_17_SAMPLES\n"); - sss_modes.push_back("\n#define USE_25_SAMPLES\n"); + { + Vector<String> sss_modes; + sss_modes.push_back("\n#define USE_11_SAMPLES\n"); + sss_modes.push_back("\n#define USE_17_SAMPLES\n"); + sss_modes.push_back("\n#define USE_25_SAMPLES\n"); - sss.shader.initialize(sss_modes); + sss.shader.initialize(sss_modes); - sss.shader_version = sss.shader.version_create(); + sss.shader_version = sss.shader.version_create(); - for (int i = 0; i < sss_modes.size(); i++) { - sss.pipelines[i] = RD::get_singleton()->compute_pipeline_create(sss.shader.version_get_shader(sss.shader_version, i)); + for (int i = 0; i < sss_modes.size(); i++) { + sss.pipelines[i] = RD::get_singleton()->compute_pipeline_create(sss.shader.version_get_shader(sss.shader_version, i)); + } } - } - { - Vector<String> resolve_modes; - resolve_modes.push_back("\n#define MODE_RESOLVE_GI\n"); - resolve_modes.push_back("\n#define MODE_RESOLVE_GI\n#define VOXEL_GI_RESOLVE\n"); + { + Vector<String> resolve_modes; + resolve_modes.push_back("\n#define MODE_RESOLVE_GI\n"); + resolve_modes.push_back("\n#define MODE_RESOLVE_GI\n#define VOXEL_GI_RESOLVE\n"); + resolve_modes.push_back("\n#define MODE_RESOLVE_DEPTH\n"); - resolve.shader.initialize(resolve_modes); + resolve.shader.initialize(resolve_modes); - resolve.shader_version = resolve.shader.version_create(); + resolve.shader_version = resolve.shader.version_create(); - for (int i = 0; i < RESOLVE_MODE_MAX; i++) { - resolve.pipelines[i] = RD::get_singleton()->compute_pipeline_create(resolve.shader.version_get_shader(resolve.shader_version, i)); + for (int i = 0; i < RESOLVE_MODE_MAX; i++) { + resolve.pipelines[i] = RD::get_singleton()->compute_pipeline_create(resolve.shader.version_get_shader(resolve.shader_version, i)); + } } } @@ -2006,30 +2526,41 @@ EffectsRD::~EffectsRD() { RD::get_singleton()->free(index_buffer); //array gets freed as dependency RD::get_singleton()->free(filter.coefficient_buffer); - RD::get_singleton()->free(ssao.mirror_sampler); - RD::get_singleton()->free(ssao.gather_constants_buffer); - RD::get_singleton()->free(ssao.importance_map_load_counter); - - bokeh.shader.version_free(bokeh.shader_version); - copy.shader.version_free(copy.shader_version); + if (prefer_raster_effects) { + blur_raster.shader.version_free(blur_raster.shader_version); + bokeh.raster_shader.version_free(blur_raster.shader_version); + luminance_reduce_raster.shader.version_free(luminance_reduce_raster.shader_version); + roughness.raster_shader.version_free(roughness.shader_version); + cubemap_downsampler.raster_shader.version_free(cubemap_downsampler.shader_version); + filter.raster_shader.version_free(filter.shader_version); + } else { + bokeh.compute_shader.version_free(bokeh.shader_version); + luminance_reduce.shader.version_free(luminance_reduce.shader_version); + roughness.compute_shader.version_free(roughness.shader_version); + cubemap_downsampler.compute_shader.version_free(cubemap_downsampler.shader_version); + filter.compute_shader.version_free(filter.shader_version); + } + if (!prefer_raster_effects) { + copy.shader.version_free(copy.shader_version); + resolve.shader.version_free(resolve.shader_version); + specular_merge.shader.version_free(specular_merge.shader_version); + ssao.blur_shader.version_free(ssao.blur_shader_version); + ssao.gather_shader.version_free(ssao.gather_shader_version); + ssao.downsample_shader.version_free(ssao.downsample_shader_version); + ssao.interleave_shader.version_free(ssao.interleave_shader_version); + ssao.importance_map_shader.version_free(ssao.importance_map_shader_version); + roughness_limiter.shader.version_free(roughness_limiter.shader_version); + ssr.shader.version_free(ssr.shader_version); + ssr_filter.shader.version_free(ssr_filter.shader_version); + ssr_scale.shader.version_free(ssr_scale.shader_version); + sss.shader.version_free(sss.shader_version); + + RD::get_singleton()->free(ssao.mirror_sampler); + RD::get_singleton()->free(ssao.gather_constants_buffer); + RD::get_singleton()->free(ssao.importance_map_load_counter); + } copy_to_fb.shader.version_free(copy_to_fb.shader_version); cube_to_dp.shader.version_free(cube_to_dp.shader_version); - cubemap_downsampler.shader.version_free(cubemap_downsampler.shader_version); - filter.shader.version_free(filter.shader_version); - luminance_reduce.shader.version_free(luminance_reduce.shader_version); - resolve.shader.version_free(resolve.shader_version); - roughness.shader.version_free(roughness.shader_version); - roughness_limiter.shader.version_free(roughness_limiter.shader_version); sort.shader.version_free(sort.shader_version); - specular_merge.shader.version_free(specular_merge.shader_version); - ssao.blur_shader.version_free(ssao.blur_shader_version); - ssao.gather_shader.version_free(ssao.gather_shader_version); - ssao.downsample_shader.version_free(ssao.downsample_shader_version); - ssao.interleave_shader.version_free(ssao.interleave_shader_version); - ssao.importance_map_shader.version_free(ssao.importance_map_shader_version); - ssr.shader.version_free(ssr.shader_version); - ssr_filter.shader.version_free(ssr_filter.shader_version); - ssr_scale.shader.version_free(ssr_scale.shader_version); - sss.shader.version_free(sss.shader_version); tonemap.shader.version_free(tonemap.shader_version); } diff --git a/servers/rendering/renderer_rd/effects_rd.h b/servers/rendering/renderer_rd/effects_rd.h index cd106e6630..0db0919dbc 100644 --- a/servers/rendering/renderer_rd/effects_rd.h +++ b/servers/rendering/renderer_rd/effects_rd.h @@ -33,14 +33,20 @@ #include "core/math/camera_matrix.h" #include "servers/rendering/renderer_rd/pipeline_cache_rd.h" +#include "servers/rendering/renderer_rd/shaders/blur_raster.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/bokeh_dof.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/bokeh_dof_raster.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/copy.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/copy_to_fb.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/cube_to_dp.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/cubemap_downsampler.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/cubemap_downsampler_raster.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/cubemap_filter.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/cubemap_filter_raster.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/cubemap_roughness.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/cubemap_roughness_raster.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/luminance_reduce.glsl.gen.h" +#include "servers/rendering/renderer_rd/shaders/luminance_reduce_raster.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/resolve.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/roughness_limiter.glsl.gen.h" #include "servers/rendering/renderer_rd/shaders/screen_space_reflection.glsl.gen.h" @@ -60,6 +66,50 @@ #include "servers/rendering_server.h" class EffectsRD { +private: + bool prefer_raster_effects; + + enum BlurRasterMode { + BLUR_MIPMAP, + + BLUR_MODE_GAUSSIAN_BLUR, + BLUR_MODE_GAUSSIAN_GLOW, + BLUR_MODE_GAUSSIAN_GLOW_AUTO_EXPOSURE, + BLUR_MODE_COPY, + + BLUR_MODE_MAX + }; + + enum { + BLUR_FLAG_HORIZONTAL = (1 << 0), + BLUR_FLAG_USE_ORTHOGONAL_PROJECTION = (1 << 1), + BLUR_FLAG_GLOW_FIRST_PASS = (1 << 2), + }; + + struct BlurRasterPushConstant { + float pixel_size[2]; + uint32_t flags; + uint32_t pad; + + //glow + float glow_strength; + float glow_bloom; + float glow_hdr_threshold; + float glow_hdr_scale; + + float glow_exposure; + float glow_white; + float glow_luminance_cap; + float glow_auto_exposure_grey; + }; + + struct BlurRaster { + BlurRasterPushConstant push_constant; + BlurRasterShaderRD shader; + RID shader_version; + PipelineCacheRD pipelines[BLUR_MODE_MAX]; + } blur_raster; + enum CopyMode { COPY_MODE_GAUSSIAN_COPY, COPY_MODE_GAUSSIAN_COPY_8BIT, @@ -161,9 +211,11 @@ class EffectsRD { struct CubemapRoughness { CubemapRoughnessPushConstant push_constant; - CubemapRoughnessShaderRD shader; + CubemapRoughnessShaderRD compute_shader; + CubemapRoughnessRasterShaderRD raster_shader; RID shader_version; - RID pipeline; + RID compute_pipeline; + PipelineCacheRD raster_pipeline; } roughness; enum TonemapMode { @@ -171,39 +223,43 @@ class EffectsRD { TONEMAP_MODE_BICUBIC_GLOW_FILTER, TONEMAP_MODE_1D_LUT, TONEMAP_MODE_BICUBIC_GLOW_FILTER_1D_LUT, + TONEMAP_MODE_SUBPASS, + TONEMAP_MODE_SUBPASS_1D_LUT, TONEMAP_MODE_NORMAL_MULTIVIEW, TONEMAP_MODE_BICUBIC_GLOW_FILTER_MULTIVIEW, TONEMAP_MODE_1D_LUT_MULTIVIEW, TONEMAP_MODE_BICUBIC_GLOW_FILTER_1D_LUT_MULTIVIEW, + TONEMAP_MODE_SUBPASS_MULTIVIEW, + TONEMAP_MODE_SUBPASS_1D_LUT_MULTIVIEW, TONEMAP_MODE_MAX }; struct TonemapPushConstant { - float bcs[3]; - uint32_t use_bcs; + float bcs[3]; // 12 - 12 + uint32_t use_bcs; // 4 - 16 - uint32_t use_glow; - uint32_t use_auto_exposure; - uint32_t use_color_correction; - uint32_t tonemapper; + uint32_t use_glow; // 4 - 20 + uint32_t use_auto_exposure; // 4 - 24 + uint32_t use_color_correction; // 4 - 28 + uint32_t tonemapper; // 4 - 32 - uint32_t glow_texture_size[2]; - float glow_intensity; - uint32_t pad3; + uint32_t glow_texture_size[2]; // 8 - 40 + float glow_intensity; // 4 - 44 + uint32_t pad3; // 4 - 48 - uint32_t glow_mode; - float glow_levels[7]; + uint32_t glow_mode; // 4 - 52 + float glow_levels[7]; // 28 - 80 - float exposure; - float white; - float auto_exposure_grey; - uint32_t pad2; + float exposure; // 4 - 84 + float white; // 4 - 88 + float auto_exposure_grey; // 4 - 92 + float luminance_multiplier; // 4 - 96 - float pixel_size[2]; - uint32_t use_fxaa; - uint32_t use_debanding; + float pixel_size[2]; // 8 - 104 + uint32_t use_fxaa; // 4 - 108 + uint32_t use_debanding; // 4 - 112 }; /* tonemap actually writes to a framebuffer, which is @@ -239,11 +295,33 @@ class EffectsRD { RID pipelines[LUMINANCE_REDUCE_MAX]; } luminance_reduce; + enum LuminanceReduceRasterMode { + LUMINANCE_REDUCE_FRAGMENT_FIRST, + LUMINANCE_REDUCE_FRAGMENT, + LUMINANCE_REDUCE_FRAGMENT_FINAL, + LUMINANCE_REDUCE_FRAGMENT_MAX + }; + + struct LuminanceReduceRasterPushConstant { + int32_t source_size[2]; + int32_t dest_size[2]; + float exposure_adjust; + float min_luminance; + float max_luminance; + uint32_t pad1; + }; + + struct LuminanceReduceFragment { + LuminanceReduceRasterPushConstant push_constant; + LuminanceReduceRasterShaderRD shader; + RID shader_version; + PipelineCacheRD pipelines[LUMINANCE_REDUCE_FRAGMENT_MAX]; + } luminance_reduce_raster; + struct CopyToDPPushConstant { float z_far; float z_near; - uint32_t z_flip; - uint32_t pad; + float texel_size[2]; float screen_rect[4]; }; @@ -281,7 +359,9 @@ class EffectsRD { enum BokehMode { BOKEH_GEN_BLUR_SIZE, BOKEH_GEN_BOKEH_BOX, + BOKEH_GEN_BOKEH_BOX_NOWEIGHT, BOKEH_GEN_BOKEH_HEXAGONAL, + BOKEH_GEN_BOKEH_HEXAGONAL_NOWEIGHT, BOKEH_GEN_BOKEH_CIRCULAR, BOKEH_COMPOSITE, BOKEH_MAX @@ -289,9 +369,11 @@ class EffectsRD { struct Bokeh { BokehPushConstant push_constant; - BokehDofShaderRD shader; + BokehDofShaderRD compute_shader; + BokehDofRasterShaderRD raster_shader; RID shader_version; - RID pipelines[BOKEH_MAX]; + RID compute_pipelines[BOKEH_MAX]; + PipelineCacheRD raster_pipelines[BOKEH_MAX]; } bokeh; enum SSAOMode { @@ -385,12 +467,10 @@ class EffectsRD { SSAODownsamplePushConstant downsample_push_constant; SsaoDownsampleShaderRD downsample_shader; RID downsample_shader_version; - RID downsample_uniform_set; SSAOGatherPushConstant gather_push_constant; SsaoShaderRD gather_shader; RID gather_shader_version; - RID gather_uniform_set; RID gather_constants_buffer; bool gather_initialized = false; @@ -398,7 +478,6 @@ class EffectsRD { SsaoImportanceMapShaderRD importance_map_shader; RID importance_map_shader_version; RID importance_map_load_counter; - RID importance_map_uniform_set; RID counter_uniform_set; SSAOBlurPushConstant blur_push_constant; @@ -429,15 +508,17 @@ class EffectsRD { struct CubemapDownsamplerPushConstant { uint32_t face_size; - float pad[3]; + uint32_t face_id; + float pad[2]; }; struct CubemapDownsampler { CubemapDownsamplerPushConstant push_constant; - CubemapDownsamplerShaderRD shader; + CubemapDownsamplerShaderRD compute_shader; + CubemapDownsamplerRasterShaderRD raster_shader; RID shader_version; - RID pipeline; - + RID compute_pipeline; + PipelineCacheRD raster_pipeline; } cubemap_downsampler; enum CubemapFilterMode { @@ -448,10 +529,19 @@ class EffectsRD { FILTER_MODE_MAX, }; + struct CubemapFilterRasterPushConstant { + uint32_t mip_level; + uint32_t face_id; + float pad[2]; + }; + struct CubemapFilter { - CubemapFilterShaderRD shader; + CubemapFilterShaderRD compute_shader; + CubemapFilterRasterShaderRD raster_shader; RID shader_version; - RID pipelines[FILTER_MODE_MAX]; + RID compute_pipelines[FILTER_MODE_MAX]; + PipelineCacheRD raster_pipelines[FILTER_MODE_MAX]; + RID uniform_set; RID image_uniform_set; RID coefficient_buffer; @@ -459,15 +549,6 @@ class EffectsRD { } filter; - struct SkyPushConstant { - float orientation[12]; // 48 - 48 - float projections[RendererSceneRender::MAX_RENDER_VIEWS][4]; // 2 x 16 - 64, if we ever need more then 3 we should consider adding this to a set. - float position[3]; // 12 - 92 - float multiplier; // 4 - 96 - float time; // 4 - 100 - float pad[3]; // 12 - 112 - }; - enum SpecularMergeMode { SPECULAR_MERGE_ADD, SPECULAR_MERGE_SSR, @@ -593,6 +674,7 @@ class EffectsRD { enum ResolveMode { RESOLVE_MODE_GI, RESOLVE_MODE_GI_VOXEL_GI, + RESOLVE_MODE_DEPTH, RESOLVE_MODE_MAX }; @@ -628,6 +710,7 @@ class EffectsRD { RID index_array; Map<RID, RID> texture_to_uniform_set_cache; + Map<RID, RID> input_to_uniform_set_cache; Map<RID, RID> image_to_uniform_set_cache; @@ -661,6 +744,7 @@ class EffectsRD { Map<TextureSamplerPair, RID> texture_sampler_to_compute_uniform_set_cache; RID _get_uniform_set_from_image(RID p_texture); + RID _get_uniform_set_for_input(RID p_texture); RID _get_uniform_set_from_texture(RID p_texture, bool p_use_mipmaps = false); RID _get_compute_uniform_set_from_texture(RID p_texture, bool p_use_mipmaps = false); RID _get_compute_uniform_set_from_texture_and_sampler(RID p_texture, RID p_sampler); @@ -668,6 +752,8 @@ class EffectsRD { RID _get_compute_uniform_set_from_image_pair(RID p_texture, RID p_texture2); public: + bool get_prefer_raster_effects(); + void copy_to_fb_rect(RID p_source_rd_texture, RID p_dest_framebuffer, const Rect2i &p_rect, bool p_flip_y = false, bool p_force_luminance = false, bool p_alpha_to_zero = false, bool p_srgb = false, RID p_secondary = RID()); void copy_to_rect(RID p_source_rd_texture, RID p_dest_texture, const Rect2i &p_rect, bool p_flip_y = false, bool p_force_luminance = false, bool p_all_source = false, bool p_8_bit_dst = false, bool p_alpha_to_one = false); void copy_cubemap_to_panorama(RID p_source_cube, RID p_dest_panorama, const Size2i &p_panorama_size, float p_lod, bool p_is_array); @@ -677,12 +763,36 @@ public: void gaussian_blur(RID p_source_rd_texture, RID p_texture, RID p_back_texture, const Rect2i &p_region, bool p_8bit_dst = false); void set_color(RID p_dest_texture, const Color &p_color, const Rect2i &p_region, bool p_8bit_dst = false); void gaussian_glow(RID p_source_rd_texture, RID p_back_texture, const Size2i &p_size, float p_strength = 1.0, bool p_high_quality = false, bool p_first_pass = false, float p_luminance_cap = 16.0, float p_exposure = 1.0, float p_bloom = 0.0, float p_hdr_bleed_treshold = 1.0, float p_hdr_bleed_scale = 1.0, RID p_auto_exposure = RID(), float p_auto_exposure_grey = 1.0); + void gaussian_glow_raster(RID p_source_rd_texture, RID p_framebuffer_half, RID p_rd_texture_half, RID p_dest_framebuffer, const Vector2 &p_pixel_size, float p_strength = 1.0, bool p_high_quality = false, bool p_first_pass = false, float p_luminance_cap = 16.0, float p_exposure = 1.0, float p_bloom = 0.0, float p_hdr_bleed_treshold = 1.0, float p_hdr_bleed_scale = 1.0, RID p_auto_exposure = RID(), float p_auto_exposure_grey = 1.0); - void cubemap_roughness(RID p_source_rd_texture, RID p_dest_framebuffer, uint32_t p_face_id, uint32_t p_sample_count, float p_roughness, float p_size); + void cubemap_roughness(RID p_source_rd_texture, RID p_dest_texture, uint32_t p_face_id, uint32_t p_sample_count, float p_roughness, float p_size); + void cubemap_roughness_raster(RID p_source_rd_texture, RID p_dest_framebuffer, uint32_t p_face_id, uint32_t p_sample_count, float p_roughness, float p_size); void make_mipmap(RID p_source_rd_texture, RID p_dest_texture, const Size2i &p_size); - void copy_cubemap_to_dp(RID p_source_rd_texture, RID p_dest_texture, const Rect2 &p_rect, float p_z_near, float p_z_far, bool p_dp_flip); + void make_mipmap_raster(RID p_source_rd_texture, RID p_dest_framebuffer, const Size2i &p_size); + void copy_cubemap_to_dp(RID p_source_rd_texture, RID p_dst_framebuffer, const Rect2 &p_rect, const Vector2 &p_dst_size, float p_z_near, float p_z_far, bool p_dp_flip); void luminance_reduction(RID p_source_texture, const Size2i p_source_size, const Vector<RID> p_reduce, RID p_prev_luminance, float p_min_luminance, float p_max_luminance, float p_adjust, bool p_set = false); - void bokeh_dof(RID p_base_texture, RID p_depth_texture, const Size2i &p_base_texture_size, RID p_secondary_texture, RID p_bokeh_texture1, RID p_bokeh_texture2, bool p_dof_far, float p_dof_far_begin, float p_dof_far_size, bool p_dof_near, float p_dof_near_begin, float p_dof_near_size, float p_bokeh_size, RS::DOFBokehShape p_bokeh_shape, RS::DOFBlurQuality p_quality, bool p_use_jitter, float p_cam_znear, float p_cam_zfar, bool p_cam_orthogonal); + void luminance_reduction_raster(RID p_source_texture, const Size2i p_source_size, const Vector<RID> p_reduce, Vector<RID> p_fb, RID p_prev_luminance, float p_min_luminance, float p_max_luminance, float p_adjust, bool p_set = false); + + struct BokehBuffers { + // bokeh buffers + + // textures + Size2i base_texture_size; + RID base_texture; + RID depth_texture; + RID secondary_texture; + RID half_texture[2]; + + // raster only + RID base_fb; + RID secondary_fb; // with weights + RID half_fb[2]; // with weights + RID base_weight_fb; + RID weight_texture[4]; + }; + + void bokeh_dof(const BokehBuffers &p_buffers, bool p_dof_far, float p_dof_far_begin, float p_dof_far_size, bool p_dof_near, float p_dof_near_begin, float p_dof_near_size, float p_bokeh_size, RS::DOFBokehShape p_bokeh_shape, RS::DOFBlurQuality p_quality, bool p_use_jitter, float p_cam_znear, float p_cam_zfar, bool p_cam_orthogonal); + void bokeh_dof_raster(const BokehBuffers &p_buffers, bool p_dof_far, float p_dof_far_begin, float p_dof_far_size, bool p_dof_near, float p_dof_near_begin, float p_dof_near_size, float p_dof_blur_amount, RenderingServer::DOFBokehShape p_bokeh_shape, RS::DOFBlurQuality p_quality, float p_cam_znear, float p_cam_zfar, bool p_cam_orthogonal); struct TonemapSettings { bool use_glow = false; @@ -708,6 +818,7 @@ public: bool use_auto_exposure = false; float auto_exposure_grey = 0.5; RID exposure_texture; + float luminance_multiplier = 1.0; bool use_bcs = false; float brightness = 1.0; @@ -745,24 +856,27 @@ public: }; void tonemapper(RID p_source_color, RID p_dst_framebuffer, const TonemapSettings &p_settings); + void tonemapper(RD::DrawListID p_subpass_draw_list, RID p_source_color, RD::FramebufferFormatID p_dst_format_id, const TonemapSettings &p_settings); - void gather_ssao(RD::ComputeListID p_compute_list, const Vector<RID> p_ao_slices, const SSAOSettings &p_settings, bool p_adaptive_base_pass); - void generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_depth_mipmaps_texture, const Vector<RID> &depth_mipmaps, RID p_ao, const Vector<RID> p_ao_slices, RID p_ao_pong, const Vector<RID> p_ao_pong_slices, RID p_upscale_buffer, RID p_importance_map, RID p_importance_map_pong, const CameraMatrix &p_projection, const SSAOSettings &p_settings, bool p_invalidate_uniform_sets); + void gather_ssao(RD::ComputeListID p_compute_list, const Vector<RID> p_ao_slices, const SSAOSettings &p_settings, bool p_adaptive_base_pass, RID p_gather_uniform_set, RID p_importance_map_uniform_set); + void generate_ssao(RID p_depth_buffer, RID p_normal_buffer, RID p_depth_mipmaps_texture, const Vector<RID> &depth_mipmaps, RID p_ao, const Vector<RID> p_ao_slices, RID p_ao_pong, const Vector<RID> p_ao_pong_slices, RID p_upscale_buffer, RID p_importance_map, RID p_importance_map_pong, const CameraMatrix &p_projection, const SSAOSettings &p_settings, bool p_invalidate_uniform_sets, RID &r_downsample_uniform_set, RID &r_gather_uniform_set, RID &r_importance_map_uniform_set); void roughness_limit(RID p_source_normal, RID p_roughness, const Size2i &p_size, float p_curve); void cubemap_downsample(RID p_source_cubemap, RID p_dest_cubemap, const Size2i &p_size); + void cubemap_downsample_raster(RID p_source_cubemap, RID p_dest_framebuffer, uint32_t p_face_id, const Size2i &p_size); void cubemap_filter(RID p_source_cubemap, Vector<RID> p_dest_cubemap, bool p_use_array); - void render_sky(RD::DrawListID p_list, float p_time, RID p_fb, RID p_samplers, RID p_fog, PipelineCacheRD *p_pipeline, RID p_uniform_set, RID p_texture_set, uint32_t p_view_count, const CameraMatrix *p_projections, const Basis &p_orientation, float p_multiplier, const Vector3 &p_position); + void cubemap_filter_raster(RID p_source_cubemap, RID p_dest_framebuffer, uint32_t p_face_id, uint32_t p_mip_level); void screen_space_reflection(RID p_diffuse, RID p_normal_roughness, RS::EnvironmentSSRRoughnessQuality p_roughness_quality, RID p_blur_radius, RID p_blur_radius2, RID p_metallic, const Color &p_metallic_mask, RID p_depth, RID p_scale_depth, RID p_scale_normal, RID p_output, RID p_output_blur, const Size2i &p_screen_size, int p_max_steps, float p_fade_in, float p_fade_out, float p_tolerance, const CameraMatrix &p_camera); void merge_specular(RID p_dest_framebuffer, RID p_specular, RID p_base, RID p_reflection); void sub_surface_scattering(RID p_diffuse, RID p_diffuse2, RID p_depth, const CameraMatrix &p_camera, const Size2i &p_screen_size, float p_scale, float p_depth_scale, RS::SubSurfaceScatteringQuality p_quality); void resolve_gi(RID p_source_depth, RID p_source_normal_roughness, RID p_source_voxel_gi, RID p_dest_depth, RID p_dest_normal_roughness, RID p_dest_voxel_gi, Vector2i p_screen_size, int p_samples, uint32_t p_barrier = RD::BARRIER_MASK_ALL); + void resolve_depth(RID p_source_depth, RID p_dest_depth, Vector2i p_screen_size, int p_samples, uint32_t p_barrier = RD::BARRIER_MASK_ALL); void sort_buffer(RID p_uniform_set, int p_size); - EffectsRD(); + EffectsRD(bool p_prefer_raster_effects); ~EffectsRD(); }; diff --git a/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.cpp b/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.cpp index 1653453c5c..1b730567d9 100644 --- a/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.cpp +++ b/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.cpp @@ -183,7 +183,7 @@ void RenderForwardClustered::RenderBufferDataForwardClustered::clear() { } } -void RenderForwardClustered::RenderBufferDataForwardClustered::configure(RID p_color_buffer, RID p_depth_buffer, int p_width, int p_height, RS::ViewportMSAA p_msaa, uint32_t p_view_count) { +void RenderForwardClustered::RenderBufferDataForwardClustered::configure(RID p_color_buffer, RID p_depth_buffer, RID p_target_buffer, int p_width, int p_height, RS::ViewportMSAA p_msaa, uint32_t p_view_count) { clear(); ERR_FAIL_COND_MSG(p_view_count != 1, "Multiple views is currently not supported in this renderer, please use the mobile renderer for VR support"); @@ -318,11 +318,11 @@ void RenderForwardClustered::_render_list_template(RenderingDevice::DrawListID p RID prev_pipeline_rd; RID prev_xforms_uniform_set; - bool shadow_pass = (p_params->pass_mode == PASS_MODE_SHADOW) || (p_params->pass_mode == PASS_MODE_SHADOW_DP); + bool shadow_pass = (p_pass_mode == PASS_MODE_SHADOW) || (p_pass_mode == PASS_MODE_SHADOW_DP); SceneState::PushConstant push_constant; - if (p_params->pass_mode == PASS_MODE_DEPTH_MATERIAL) { + if (p_pass_mode == PASS_MODE_DEPTH_MATERIAL) { push_constant.uv_offset = Math::make_half_float(p_params->uv_offset.y) << 16; push_constant.uv_offset |= Math::make_half_float(p_params->uv_offset.x); } else { @@ -339,14 +339,26 @@ void RenderForwardClustered::_render_list_template(RenderingDevice::DrawListID p SceneShaderForwardClustered::ShaderData *shader; void *mesh_surface; - if (shadow_pass || p_params->pass_mode == PASS_MODE_DEPTH) { //regular depth pass can use these too + if (shadow_pass || p_pass_mode == PASS_MODE_DEPTH) { //regular depth pass can use these too material_uniform_set = surf->material_uniform_set_shadow; shader = surf->shader_shadow; mesh_surface = surf->surface_shadow; } else { - material_uniform_set = surf->material_uniform_set; - shader = surf->shader; +#ifdef DEBUG_ENABLED + if (unlikely(get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_LIGHTING)) { + material_uniform_set = scene_shader.default_material_uniform_set; + shader = scene_shader.default_material_shader_ptr; + } else if (unlikely(get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_OVERDRAW)) { + material_uniform_set = scene_shader.overdraw_material_uniform_set; + shader = scene_shader.overdraw_material_shader_ptr; + } else { +#endif + material_uniform_set = surf->material_uniform_set; + shader = surf->shader; +#ifdef DEBUG_ENABLED + } +#endif mesh_surface = surf->surface; } @@ -357,7 +369,7 @@ void RenderForwardClustered::_render_list_template(RenderingDevice::DrawListID p //find cull variant SceneShaderForwardClustered::ShaderData::CullVariant cull_variant; - if (p_params->pass_mode == PASS_MODE_DEPTH_MATERIAL || p_params->pass_mode == PASS_MODE_SDF || ((p_params->pass_mode == PASS_MODE_SHADOW || p_params->pass_mode == PASS_MODE_SHADOW_DP) && surf->flags & GeometryInstanceSurfaceDataCache::FLAG_USES_DOUBLE_SIDED_SHADOWS)) { + if (p_pass_mode == PASS_MODE_DEPTH_MATERIAL || p_pass_mode == PASS_MODE_SDF || ((p_pass_mode == PASS_MODE_SHADOW || p_pass_mode == PASS_MODE_SHADOW_DP) && surf->flags & GeometryInstanceSurfaceDataCache::FLAG_USES_DOUBLE_SIDED_SHADOWS)) { cull_variant = SceneShaderForwardClustered::ShaderData::CULL_VARIANT_DOUBLE_SIDED; } else { bool mirror = surf->owner->mirror; @@ -372,14 +384,30 @@ void RenderForwardClustered::_render_list_template(RenderingDevice::DrawListID p SceneShaderForwardClustered::ShaderVersion shader_version = SceneShaderForwardClustered::SHADER_VERSION_MAX; // Assigned to silence wrong -Wmaybe-initialized. - switch (p_params->pass_mode) { + uint32_t pipeline_specialization = 0; + + if (p_pass_mode == PASS_MODE_COLOR || p_pass_mode == PASS_MODE_COLOR_TRANSPARENT || p_pass_mode == PASS_MODE_COLOR_SPECULAR) { + if (element_info.uses_softshadow) { + pipeline_specialization |= SceneShaderForwardClustered::SHADER_SPECIALIZATION_SOFT_SHADOWS; + } + if (element_info.uses_projector) { + pipeline_specialization |= SceneShaderForwardClustered::SHADER_SPECIALIZATION_PROJECTOR; + } + + if (p_params->use_directional_soft_shadow) { + pipeline_specialization |= SceneShaderForwardClustered::SHADER_SPECIALIZATION_DIRECTIONAL_SOFT_SHADOWS; + } + } + + switch (p_pass_mode) { case PASS_MODE_COLOR: case PASS_MODE_COLOR_TRANSPARENT: { if (element_info.uses_lightmap) { shader_version = SceneShaderForwardClustered::SHADER_VERSION_LIGHTMAP_COLOR_PASS; - } else if (element_info.uses_forward_gi) { - shader_version = SceneShaderForwardClustered::SHADER_VERSION_COLOR_PASS_WITH_FORWARD_GI; } else { + if (element_info.uses_forward_gi) { + pipeline_specialization |= SceneShaderForwardClustered::SHADER_SPECIALIZATION_FORWARD_GI; + } shader_version = SceneShaderForwardClustered::SHADER_VERSION_COLOR_PASS; } } break; @@ -440,7 +468,7 @@ void RenderForwardClustered::_render_list_template(RenderingDevice::DrawListID p prev_index_array_rd = index_array_rd; } - RID pipeline_rd = pipeline->get_render_pipeline(vertex_format, framebuffer_format, p_params->force_wireframe); + RID pipeline_rd = pipeline->get_render_pipeline(vertex_format, framebuffer_format, p_params->force_wireframe, 0, pipeline_specialization); if (pipeline_rd != prev_pipeline_rd) { // checking with prev shader does not make so much sense, as @@ -455,8 +483,8 @@ void RenderForwardClustered::_render_list_template(RenderingDevice::DrawListID p } if (material_uniform_set != prev_material_uniform_set) { - //update uniform set - if (material_uniform_set.is_valid()) { + // Update uniform set. + if (material_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(material_uniform_set)) { // Material may not have a uniform set. RD::get_singleton()->draw_list_bind_uniform_set(draw_list, material_uniform_set, MATERIAL_UNIFORM_SET); } @@ -561,11 +589,6 @@ void RenderForwardClustered::_setup_environment(const RenderDataRD *p_render_dat RendererStorageRD::store_soft_shadow_kernel(penumbra_shadow_kernel_get(), scene_state.ubo.penumbra_shadow_kernel); RendererStorageRD::store_soft_shadow_kernel(soft_shadow_kernel_get(), scene_state.ubo.soft_shadow_kernel); - scene_state.ubo.directional_penumbra_shadow_samples = directional_penumbra_shadow_samples_get(); - scene_state.ubo.directional_soft_shadow_samples = directional_soft_shadow_samples_get(); - scene_state.ubo.penumbra_shadow_samples = penumbra_shadow_samples_get(); - scene_state.ubo.soft_shadow_samples = soft_shadow_samples_get(); - Size2 screen_pixel_size = Vector2(1.0, 1.0) / Size2(p_screen_size); scene_state.ubo.screen_pixel_size[0] = screen_pixel_size.x; scene_state.ubo.screen_pixel_size[1] = screen_pixel_size.y; @@ -801,13 +824,16 @@ void RenderForwardClustered::_update_instance_data_buffer(RenderListType p_rende RD::get_singleton()->buffer_update(scene_state.instance_buffer[p_render_list], 0, sizeof(SceneState::InstanceData) * scene_state.instance_data[p_render_list].size(), scene_state.instance_data[p_render_list].ptr(), RD::BARRIER_MASK_RASTER); } } -void RenderForwardClustered::_fill_instance_data(RenderListType p_render_list, uint32_t p_offset, int32_t p_max_elements, bool p_update_buffer) { +void RenderForwardClustered::_fill_instance_data(RenderListType p_render_list, int *p_render_info, uint32_t p_offset, int32_t p_max_elements, bool p_update_buffer) { RenderList *rl = &render_list[p_render_list]; uint32_t element_total = p_max_elements >= 0 ? uint32_t(p_max_elements) : rl->elements.size(); scene_state.instance_data[p_render_list].resize(p_offset + element_total); rl->element_info.resize(p_offset + element_total); + if (p_render_info) { + p_render_info[RS::VIEWPORT_RENDER_INFO_OBJECTS_IN_FRAME] += element_total; + } uint32_t repeats = 0; GeometryInstanceSurfaceDataCache *prev_surface = nullptr; for (uint32_t i = 0; i < element_total; i++) { @@ -833,7 +859,7 @@ void RenderForwardClustered::_fill_instance_data(RenderListType p_render_list, u bool cant_repeat = instance_data.flags & INSTANCE_DATA_FLAG_MULTIMESH || inst->mesh_instance.is_valid(); - if (prev_surface != nullptr && !cant_repeat && prev_surface->sort.sort_key1 == surface->sort.sort_key1 && prev_surface->sort.sort_key2 == surface->sort.sort_key2) { + if (prev_surface != nullptr && !cant_repeat && prev_surface->sort.sort_key1 == surface->sort.sort_key1 && prev_surface->sort.sort_key2 == surface->sort.sort_key2 && repeats < RenderElementInfo::MAX_REPEATS) { //this element is the same as the previous one, count repeats to draw it using instancing repeats++; } else { @@ -843,6 +869,9 @@ void RenderForwardClustered::_fill_instance_data(RenderListType p_render_list, u } } repeats = 1; + if (p_render_info) { + p_render_info[RS::VIEWPORT_RENDER_INFO_DRAW_CALLS_IN_FRAME]++; + } } RenderElementInfo &element_info = rl->element_info[p_offset + i]; @@ -850,6 +879,8 @@ void RenderForwardClustered::_fill_instance_data(RenderListType p_render_list, u element_info.lod_index = surface->sort.lod_index; element_info.uses_forward_gi = surface->sort.uses_forward_gi; element_info.uses_lightmap = surface->sort.uses_lightmap; + element_info.uses_softshadow = surface->sort.uses_softshadow; + element_info.uses_projector = surface->sort.uses_projector; if (cant_repeat) { prev_surface = nullptr; @@ -869,6 +900,11 @@ void RenderForwardClustered::_fill_instance_data(RenderListType p_render_list, u } } +_FORCE_INLINE_ static uint32_t _indices_to_primitives(RS::PrimitiveType p_primitive, uint32_t p_indices) { + static const uint32_t divisor[RS::PRIMITIVE_MAX] = { 1, 2, 1, 3, 1 }; + static const uint32_t subtractor[RS::PRIMITIVE_MAX] = { 0, 0, 1, 0, 1 }; + return (p_indices - subtractor[p_primitive]) / divisor[p_primitive]; +} void RenderForwardClustered::_fill_render_list(RenderListType p_render_list, const RenderDataRD *p_render_data, PassMode p_pass_mode, bool p_using_sdfgi, bool p_using_opaque_gi, bool p_append) { if (p_render_list == RENDER_LIST_OPAQUE) { scene_state.used_sss = false; @@ -903,6 +939,9 @@ void RenderForwardClustered::_fill_render_list(RenderListType p_render_list, con uint32_t flags = inst->base_flags; //fill flags if appropriate + if (inst->non_uniform_scale) { + flags |= INSTANCE_DATA_FLAGS_NON_UNIFORM_SCALE; + } bool uses_lightmap = false; bool uses_gi = false; @@ -1008,17 +1047,41 @@ void RenderForwardClustered::_fill_render_list(RenderListType p_render_list, con distance = -distance_max; } - surf->sort.lod_index = storage->mesh_surface_get_lod(surf->surface, inst->lod_model_scale * inst->lod_bias, distance * p_render_data->lod_distance_multiplier, p_render_data->screen_lod_threshold); + uint32_t indices; + surf->sort.lod_index = storage->mesh_surface_get_lod(surf->surface, inst->lod_model_scale * inst->lod_bias, distance * p_render_data->lod_distance_multiplier, p_render_data->screen_lod_threshold, &indices); + if (p_render_data->render_info) { + indices = _indices_to_primitives(surf->primitive, indices); + if (p_render_list == RENDER_LIST_OPAQUE) { //opaque + p_render_data->render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_VISIBLE][RS::VIEWPORT_RENDER_INFO_PRIMITIVES_IN_FRAME] += indices; + } else if (p_render_list == RENDER_LIST_SECONDARY) { //shadow + p_render_data->render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_SHADOW][RS::VIEWPORT_RENDER_INFO_PRIMITIVES_IN_FRAME] += indices; + } + } } else { surf->sort.lod_index = 0; + if (p_render_data->render_info) { + uint32_t to_draw = storage->mesh_surface_get_vertices_drawn_count(surf->surface); + to_draw = _indices_to_primitives(surf->primitive, to_draw); + to_draw *= inst->instance_count; + if (p_render_list == RENDER_LIST_OPAQUE) { //opaque + p_render_data->render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_VISIBLE][RS::VIEWPORT_RENDER_INFO_PRIMITIVES_IN_FRAME] += storage->mesh_surface_get_vertices_drawn_count(surf->surface); + } else if (p_render_list == RENDER_LIST_SECONDARY) { //shadow + p_render_data->render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_SHADOW][RS::VIEWPORT_RENDER_INFO_PRIMITIVES_IN_FRAME] += storage->mesh_surface_get_vertices_drawn_count(surf->surface); + } + } } // ADD Element if (p_pass_mode == PASS_MODE_COLOR) { - if (surf->flags & (GeometryInstanceSurfaceDataCache::FLAG_PASS_DEPTH | GeometryInstanceSurfaceDataCache::FLAG_PASS_OPAQUE)) { +#ifdef DEBUG_ENABLED + bool force_alpha = unlikely(get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_OVERDRAW); +#else + bool force_alpha = false; +#endif + if (!force_alpha && (surf->flags & (GeometryInstanceSurfaceDataCache::FLAG_PASS_DEPTH | GeometryInstanceSurfaceDataCache::FLAG_PASS_OPAQUE))) { rl->add_element(surf); } - if (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_PASS_ALPHA) { + if (force_alpha || (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_PASS_ALPHA)) { render_list[RENDER_LIST_ALPHA].add_element(surf); if (uses_gi) { surf->sort.uses_forward_gi = 1; @@ -1100,6 +1163,7 @@ void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Co render_buffer = (RenderBufferDataForwardClustered *)render_buffers_get_data(p_render_data->render_buffers); } RendererSceneEnvironmentRD *env = get_environment(p_render_data->environment); + static const int texture_multisamples[RS::VIEWPORT_MSAA_MAX] = { 1, 2, 4, 8, 16 }; //first of all, make a new render pass //fill up ubo @@ -1209,8 +1273,8 @@ void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Co _fill_render_list(RENDER_LIST_OPAQUE, p_render_data, PASS_MODE_COLOR, using_sdfgi, using_sdfgi || using_voxelgi); render_list[RENDER_LIST_OPAQUE].sort_by_key(); - render_list[RENDER_LIST_ALPHA].sort_by_depth(); - _fill_instance_data(RENDER_LIST_OPAQUE); + render_list[RENDER_LIST_ALPHA].sort_by_reverse_depth_and_priority(); + _fill_instance_data(RENDER_LIST_OPAQUE, p_render_data->render_info ? p_render_data->render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_VISIBLE] : (int *)nullptr); _fill_instance_data(RENDER_LIST_ALPHA); RD::get_singleton()->draw_command_end_label(); @@ -1324,7 +1388,7 @@ void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Co RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_OPAQUE, nullptr, RID()); bool finish_depth = using_ssao || using_sdfgi || using_voxelgi; - RenderListParameters render_list_params(render_list[RENDER_LIST_OPAQUE].elements.ptr(), render_list[RENDER_LIST_OPAQUE].element_info.ptr(), render_list[RENDER_LIST_OPAQUE].elements.size(), reverse_cull, depth_pass_mode, render_buffer == nullptr, rp_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), p_render_data->lod_camera_plane, p_render_data->lod_distance_multiplier, p_render_data->screen_lod_threshold); + RenderListParameters render_list_params(render_list[RENDER_LIST_OPAQUE].elements.ptr(), render_list[RENDER_LIST_OPAQUE].element_info.ptr(), render_list[RENDER_LIST_OPAQUE].elements.size(), reverse_cull, depth_pass_mode, render_buffer == nullptr, p_render_data->directional_light_soft_shadows, rp_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), p_render_data->lod_camera_plane, p_render_data->lod_distance_multiplier, p_render_data->screen_lod_threshold); _render_list_with_threads(&render_list_params, depth_framebuffer, needs_pre_resolve ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, needs_pre_resolve ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_CLEAR, finish_depth ? RD::FINAL_ACTION_READ : RD::FINAL_ACTION_CONTINUE, needs_pre_resolve ? Vector<Color>() : depth_pass_clear); RD::get_singleton()->draw_command_end_label(); @@ -1340,10 +1404,9 @@ void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Co if (needs_pre_resolve) { RD::get_singleton()->barrier(RD::BARRIER_MASK_RASTER, RD::BARRIER_MASK_COMPUTE); } - static int texture_samples[RS::VIEWPORT_MSAA_MAX] = { 1, 2, 4, 8, 16 }; - storage->get_effects()->resolve_gi(render_buffer->depth_msaa, render_buffer->normal_roughness_buffer_msaa, using_voxelgi ? render_buffer->voxelgi_buffer_msaa : RID(), render_buffer->depth, render_buffer->normal_roughness_buffer, using_voxelgi ? render_buffer->voxelgi_buffer : RID(), Vector2i(render_buffer->width, render_buffer->height), texture_samples[render_buffer->msaa]); + storage->get_effects()->resolve_gi(render_buffer->depth_msaa, render_buffer->normal_roughness_buffer_msaa, using_voxelgi ? render_buffer->voxelgi_buffer_msaa : RID(), render_buffer->depth, render_buffer->normal_roughness_buffer, using_voxelgi ? render_buffer->voxelgi_buffer : RID(), Vector2i(render_buffer->width, render_buffer->height), texture_multisamples[render_buffer->msaa]); } else if (finish_depth) { - RD::get_singleton()->texture_resolve_multisample(render_buffer->depth_msaa, render_buffer->depth); + storage->get_effects()->resolve_depth(render_buffer->depth_msaa, render_buffer->depth, Vector2i(render_buffer->width, render_buffer->height), texture_multisamples[render_buffer->msaa]); } RD::get_singleton()->draw_command_end_label(); } @@ -1382,7 +1445,7 @@ void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Co } RID framebuffer = using_separate_specular ? opaque_specular_framebuffer : opaque_framebuffer; - RenderListParameters render_list_params(render_list[RENDER_LIST_OPAQUE].elements.ptr(), render_list[RENDER_LIST_OPAQUE].element_info.ptr(), render_list[RENDER_LIST_OPAQUE].elements.size(), reverse_cull, using_separate_specular ? PASS_MODE_COLOR_SPECULAR : PASS_MODE_COLOR, render_buffer == nullptr, rp_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), p_render_data->lod_camera_plane, p_render_data->lod_distance_multiplier, p_render_data->screen_lod_threshold); + RenderListParameters render_list_params(render_list[RENDER_LIST_OPAQUE].elements.ptr(), render_list[RENDER_LIST_OPAQUE].element_info.ptr(), render_list[RENDER_LIST_OPAQUE].elements.size(), reverse_cull, using_separate_specular ? PASS_MODE_COLOR_SPECULAR : PASS_MODE_COLOR, render_buffer == nullptr, p_render_data->directional_light_soft_shadows, rp_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), p_render_data->lod_camera_plane, p_render_data->lod_distance_multiplier, p_render_data->screen_lod_threshold); _render_list_with_threads(&render_list_params, framebuffer, keep_color ? RD::INITIAL_ACTION_KEEP : RD::INITIAL_ACTION_CLEAR, will_continue_color ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, depth_pre_pass ? (continue_depth ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_KEEP) : RD::INITIAL_ACTION_CLEAR, will_continue_depth ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, c, 1.0, 0); if (will_continue_color && using_separate_specular) { // close the specular framebuffer, as it's no longer used @@ -1447,7 +1510,7 @@ void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Co } if (render_buffer && !can_continue_depth && render_buffer->msaa != RS::VIEWPORT_MSAA_DISABLED) { - RD::get_singleton()->texture_resolve_multisample(render_buffer->depth_msaa, render_buffer->depth); + storage->get_effects()->resolve_depth(render_buffer->depth_msaa, render_buffer->depth, Vector2i(render_buffer->width, render_buffer->height), texture_multisamples[render_buffer->msaa]); } if (using_separate_specular) { @@ -1470,6 +1533,16 @@ void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Co } } + if (scene_state.used_screen_texture) { + // Copy screen texture to backbuffer so we can read from it + _render_buffers_copy_screen_texture(p_render_data); + } + + if (scene_state.used_depth_texture) { + // Copy depth texture to backbuffer so we can read from it + _render_buffers_copy_depth_texture(p_render_data); + } + RENDER_TIMESTAMP("Render Transparent Pass"); RD::get_singleton()->draw_command_begin_label("Render Transparent Pass"); @@ -1479,7 +1552,7 @@ void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Co _setup_environment(p_render_data, p_render_data->reflection_probe.is_valid(), screen_size, !p_render_data->reflection_probe.is_valid(), p_default_bg_color, false); { - RenderListParameters render_list_params(render_list[RENDER_LIST_ALPHA].elements.ptr(), render_list[RENDER_LIST_ALPHA].element_info.ptr(), render_list[RENDER_LIST_ALPHA].elements.size(), false, PASS_MODE_COLOR, render_buffer == nullptr, rp_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), p_render_data->lod_camera_plane, p_render_data->lod_distance_multiplier, p_render_data->screen_lod_threshold); + RenderListParameters render_list_params(render_list[RENDER_LIST_ALPHA].elements.ptr(), render_list[RENDER_LIST_ALPHA].element_info.ptr(), render_list[RENDER_LIST_ALPHA].elements.size(), false, PASS_MODE_COLOR, render_buffer == nullptr, p_render_data->directional_light_soft_shadows, rp_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), p_render_data->lod_camera_plane, p_render_data->lod_distance_multiplier, p_render_data->screen_lod_threshold); _render_list_with_threads(&render_list_params, alpha_framebuffer, can_continue_color ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, can_continue_depth ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ); } @@ -1492,6 +1565,14 @@ void RenderForwardClustered::_render_scene(RenderDataRD *p_render_data, const Co } RD::get_singleton()->draw_command_end_label(); + + if (p_render_data->render_buffers.is_valid()) { + _debug_draw_cluster(p_render_data->render_buffers); + + RENDER_TIMESTAMP("Tonemap"); + + _render_buffers_post_process_and_tonemap(p_render_data); + } } void RenderForwardClustered::_render_shadow_begin() { @@ -1503,7 +1584,7 @@ void RenderForwardClustered::_render_shadow_begin() { scene_state.instance_data[RENDER_LIST_SECONDARY].clear(); } -void RenderForwardClustered::_render_shadow_append(RID p_framebuffer, const PagedArray<GeometryInstance *> &p_instances, const CameraMatrix &p_projection, const Transform3D &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane, float p_lod_distance_multiplier, float p_screen_lod_threshold, const Rect2i &p_rect, bool p_flip_y, bool p_clear_region, bool p_begin, bool p_end) { +void RenderForwardClustered::_render_shadow_append(RID p_framebuffer, const PagedArray<GeometryInstance *> &p_instances, const CameraMatrix &p_projection, const Transform3D &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane, float p_lod_distance_multiplier, float p_screen_lod_threshold, const Rect2i &p_rect, bool p_flip_y, bool p_clear_region, bool p_begin, bool p_end, RendererScene::RenderInfo *p_render_info) { uint32_t shadow_pass_index = scene_state.shadow_passes.size(); SceneState::ShadowPass shadow_pass; @@ -1518,6 +1599,7 @@ void RenderForwardClustered::_render_shadow_append(RID p_framebuffer, const Page render_data.instances = &p_instances; render_data.lod_camera_plane = p_camera_plane; render_data.lod_distance_multiplier = p_lod_distance_multiplier; + render_data.render_info = p_render_info; scene_state.ubo.dual_paraboloid_side = p_use_dp_flip ? -1 : 1; @@ -1535,7 +1617,7 @@ void RenderForwardClustered::_render_shadow_append(RID p_framebuffer, const Page _fill_render_list(RENDER_LIST_SECONDARY, &render_data, pass_mode, false, false, true); uint32_t render_list_size = render_list[RENDER_LIST_SECONDARY].elements.size() - render_list_from; render_list[RENDER_LIST_SECONDARY].sort_by_key_range(render_list_from, render_list_size); - _fill_instance_data(RENDER_LIST_SECONDARY, render_list_from, render_list_size, false); + _fill_instance_data(RENDER_LIST_SECONDARY, p_render_info ? p_render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_SHADOW] : (int *)nullptr, render_list_from, render_list_size, false); { //regular forward for now @@ -1580,7 +1662,7 @@ void RenderForwardClustered::_render_shadow_end(uint32_t p_barrier) { for (uint32_t i = 0; i < scene_state.shadow_passes.size(); i++) { SceneState::ShadowPass &shadow_pass = scene_state.shadow_passes[i]; - RenderListParameters render_list_parameters(render_list[RENDER_LIST_SECONDARY].elements.ptr() + shadow_pass.element_from, render_list[RENDER_LIST_SECONDARY].element_info.ptr() + shadow_pass.element_from, shadow_pass.element_count, shadow_pass.flip_cull, shadow_pass.pass_mode, true, shadow_pass.rp_uniform_set, false, Vector2(), shadow_pass.camera_plane, shadow_pass.lod_distance_multiplier, shadow_pass.screen_lod_threshold, shadow_pass.element_from, RD::BARRIER_MASK_NO_BARRIER); + RenderListParameters render_list_parameters(render_list[RENDER_LIST_SECONDARY].elements.ptr() + shadow_pass.element_from, render_list[RENDER_LIST_SECONDARY].element_info.ptr() + shadow_pass.element_from, shadow_pass.element_count, shadow_pass.flip_cull, shadow_pass.pass_mode, true, false, shadow_pass.rp_uniform_set, false, Vector2(), shadow_pass.camera_plane, shadow_pass.lod_distance_multiplier, shadow_pass.screen_lod_threshold, shadow_pass.element_from, RD::BARRIER_MASK_NO_BARRIER); _render_list_with_threads(&render_list_parameters, shadow_pass.framebuffer, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, shadow_pass.initial_depth_action, shadow_pass.final_depth_action, Vector<Color>(), 1.0, 0, shadow_pass.rect); } @@ -1621,7 +1703,7 @@ void RenderForwardClustered::_render_particle_collider_heightfield(RID p_fb, con { //regular forward for now - RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), false, pass_mode, true, rp_uniform_set); + RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), false, pass_mode, true, false, rp_uniform_set); _render_list_with_threads(&render_list_params, p_fb, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ); } RD::get_singleton()->draw_command_end_label(); @@ -1656,7 +1738,7 @@ void RenderForwardClustered::_render_material(const Transform3D &p_cam_transform RENDER_TIMESTAMP("Render Material"); { - RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), true, pass_mode, true, rp_uniform_set); + RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), true, pass_mode, true, false, rp_uniform_set); //regular forward for now Vector<Color> clear; clear.push_back(Color(0, 0, 0, 0)); @@ -1699,7 +1781,7 @@ void RenderForwardClustered::_render_uv2(const PagedArray<GeometryInstance *> &p RENDER_TIMESTAMP("Render Material"); { - RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), true, pass_mode, true, rp_uniform_set, true); + RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), true, pass_mode, true, false, rp_uniform_set, true); //regular forward for now Vector<Color> clear; clear.push_back(Color(0, 0, 0, 0)); @@ -1817,7 +1899,7 @@ void RenderForwardClustered::_render_sdfgi(RID p_render_buffers, const Vector3i E = sdfgi_framebuffer_size_cache.insert(fb_size, fb); } - RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), true, pass_mode, true, rp_uniform_set, false); + RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), true, pass_mode, true, false, rp_uniform_set, false); _render_list_with_threads(&render_list_params, E->get(), RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, Vector<Color>(), 1.0, 0, Rect2(), sbs); } @@ -1873,13 +1955,67 @@ void RenderForwardClustered::_update_render_base_uniform_set() { { RD::Uniform u; u.binding = 3; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + RID sampler; + switch (decals_get_filter()) { + case RS::DECAL_FILTER_NEAREST: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::DECAL_FILTER_NEAREST_MIPMAPS: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::DECAL_FILTER_LINEAR: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::DECAL_FILTER_LINEAR_MIPMAPS: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::DECAL_FILTER_LINEAR_MIPMAPS_ANISOTROPIC: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + } + + u.ids.push_back(sampler); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 4; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + RID sampler; + switch (light_projectors_get_filter()) { + case RS::LIGHT_PROJECTOR_FILTER_NEAREST: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::LIGHT_PROJECTOR_FILTER_NEAREST_MIPMAPS: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::LIGHT_PROJECTOR_FILTER_LINEAR: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::LIGHT_PROJECTOR_FILTER_LINEAR_MIPMAPS: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::LIGHT_PROJECTOR_FILTER_LINEAR_MIPMAPS_ANISOTROPIC: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + } + + u.ids.push_back(sampler); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 5; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.ids.push_back(get_omni_light_buffer()); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 4; + u.binding = 6; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.ids.push_back(get_spot_light_buffer()); uniforms.push_back(u); @@ -1887,35 +2023,35 @@ void RenderForwardClustered::_update_render_base_uniform_set() { { RD::Uniform u; - u.binding = 5; + u.binding = 7; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.ids.push_back(get_reflection_probe_buffer()); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 6; + u.binding = 8; u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; u.ids.push_back(get_directional_light_buffer()); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 7; + u.binding = 9; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.ids.push_back(scene_state.lightmap_buffer); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 8; + u.binding = 10; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.ids.push_back(scene_state.lightmap_capture_buffer); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 9; + u.binding = 11; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID decal_atlas = storage->decal_atlas_get_texture(); u.ids.push_back(decal_atlas); @@ -1923,7 +2059,7 @@ void RenderForwardClustered::_update_render_base_uniform_set() { } { RD::Uniform u; - u.binding = 10; + u.binding = 12; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID decal_atlas = storage->decal_atlas_get_texture_srgb(); u.ids.push_back(decal_atlas); @@ -1931,7 +2067,7 @@ void RenderForwardClustered::_update_render_base_uniform_set() { } { RD::Uniform u; - u.binding = 11; + u.binding = 13; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.ids.push_back(get_decal_buffer()); uniforms.push_back(u); @@ -1940,7 +2076,7 @@ void RenderForwardClustered::_update_render_base_uniform_set() { { RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; - u.binding = 12; + u.binding = 14; u.ids.push_back(storage->global_variables_get_storage_buffer()); uniforms.push_back(u); } @@ -1948,7 +2084,7 @@ void RenderForwardClustered::_update_render_base_uniform_set() { { RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; - u.binding = 13; + u.binding = 15; u.ids.push_back(sdfgi_get_ubo()); uniforms.push_back(u); } @@ -2093,7 +2229,8 @@ RID RenderForwardClustered::_setup_render_pass_uniform_set(RenderListType p_rend RD::Uniform u; u.binding = 9; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - RID texture = (false && rb && rb->depth.is_valid()) ? rb->depth : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE); + RID dbt = rb ? render_buffers_get_back_depth_texture(p_render_data->render_buffers) : RID(); + RID texture = (dbt.is_valid()) ? dbt : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE); u.ids.push_back(texture); uniforms.push_back(u); } @@ -2460,12 +2597,14 @@ void RenderForwardClustered::_geometry_instance_add_surface_with_material(Geomet sdcache->sort.sort_key2 = 0; sdcache->sort.surface_index = p_surface; - sdcache->sort.material_id_low = p_material_id & 0x3FFF; - sdcache->sort.material_id_hi = p_material_id >> 14; + sdcache->sort.material_id_low = p_material_id & 0xFFFF; + sdcache->sort.material_id_hi = p_material_id >> 16; sdcache->sort.shader_id = p_shader_id; sdcache->sort.geometry_id = p_mesh.get_local_index(); //only meshes can repeat anyway sdcache->sort.uses_forward_gi = ginstance->can_sdfgi; sdcache->sort.priority = p_material->priority; + sdcache->sort.uses_projector = ginstance->using_projectors; + sdcache->sort.uses_softshadow = ginstance->using_softshadows; } void RenderForwardClustered::_geometry_instance_add_surface(GeometryInstanceForwardClustered *ginstance, uint32_t p_surface, RID p_material, RID p_mesh) { @@ -2594,6 +2733,8 @@ void RenderForwardClustered::_geometry_instance_update(GeometryInstance *p_geome //Fill push constant + ginstance->base_flags = 0; + bool store_transform = true; if (ginstance->data->base_type == RS::INSTANCE_MULTIMESH) { @@ -2737,6 +2878,7 @@ void RenderForwardClustered::geometry_instance_set_transform(GeometryInstance *p float max_scale = MAX(model_scale_vec.x, MAX(model_scale_vec.y, model_scale_vec.z)); float min_scale = MIN(model_scale_vec.x, MIN(model_scale_vec.y, model_scale_vec.z)); + ginstance->non_uniform_scale = max_scale >= 0.0 && (min_scale / max_scale) < 0.9; ginstance->lod_model_scale = max_scale; @@ -2857,6 +2999,56 @@ void RenderForwardClustered::geometry_instance_pair_voxel_gi_instances(GeometryI } } +void RenderForwardClustered::geometry_instance_set_softshadow_projector_pairing(GeometryInstance *p_geometry_instance, bool p_softshadow, bool p_projector) { + GeometryInstanceForwardClustered *ginstance = static_cast<GeometryInstanceForwardClustered *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + ginstance->using_projectors = p_projector; + ginstance->using_softshadows = p_softshadow; + _geometry_instance_mark_dirty(ginstance); +} + +void RenderForwardClustered::_update_shader_quality_settings() { + Vector<RD::PipelineSpecializationConstant> spec_constants; + + RD::PipelineSpecializationConstant sc; + sc.type = RD::PIPELINE_SPECIALIZATION_CONSTANT_TYPE_INT; + + sc.constant_id = SPEC_CONSTANT_SOFT_SHADOW_SAMPLES; + sc.int_value = soft_shadow_samples_get(); + + spec_constants.push_back(sc); + + sc.constant_id = SPEC_CONSTANT_PENUMBRA_SHADOW_SAMPLES; + sc.int_value = penumbra_shadow_samples_get(); + + spec_constants.push_back(sc); + + sc.constant_id = SPEC_CONSTANT_DIRECTIONAL_SOFT_SHADOW_SAMPLES; + sc.int_value = directional_soft_shadow_samples_get(); + + spec_constants.push_back(sc); + + sc.constant_id = SPEC_CONSTANT_DIRECTIONAL_PENUMBRA_SHADOW_SAMPLES; + sc.int_value = directional_penumbra_shadow_samples_get(); + + spec_constants.push_back(sc); + + sc.type = RD::PIPELINE_SPECIALIZATION_CONSTANT_TYPE_BOOL; + sc.constant_id = SPEC_CONSTANT_DECAL_FILTER; + sc.bool_value = decals_get_filter() == RS::DECAL_FILTER_NEAREST_MIPMAPS || decals_get_filter() == RS::DECAL_FILTER_LINEAR_MIPMAPS || decals_get_filter() == RS::DECAL_FILTER_LINEAR_MIPMAPS_ANISOTROPIC; + + spec_constants.push_back(sc); + + sc.constant_id = SPEC_CONSTANT_PROJECTOR_FILTER; + sc.bool_value = light_projectors_get_filter() == RS::LIGHT_PROJECTOR_FILTER_NEAREST_MIPMAPS || light_projectors_get_filter() == RS::LIGHT_PROJECTOR_FILTER_LINEAR_MIPMAPS || light_projectors_get_filter() == RS::LIGHT_PROJECTOR_FILTER_LINEAR_MIPMAPS_ANISOTROPIC; + + spec_constants.push_back(sc); + + scene_shader.set_default_specialization_constants(spec_constants); + + _base_uniforms_changed(); //also need this +} + RenderForwardClustered::RenderForwardClustered(RendererStorageRD *p_storage) : RendererSceneRenderRD(p_storage) { singleton = this; @@ -2870,7 +3062,7 @@ RenderForwardClustered::RenderForwardClustered(RendererStorageRD *p_storage) : defines += "\n#define USE_RADIANCE_CUBEMAP_ARRAY \n"; } defines += "\n#define SDFGI_OCT_SIZE " + itos(gi.sdfgi_get_lightprobe_octahedron_size()) + "\n"; - defines += "\n#define MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS " + itos(get_max_directional_lights()) + "\n"; + defines += "\n#define MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS " + itos(MAX_DIRECTIONAL_LIGHTS) + "\n"; { //lightmaps @@ -2894,6 +3086,8 @@ RenderForwardClustered::RenderForwardClustered(RendererStorageRD *p_storage) : } render_list_thread_threshold = GLOBAL_GET("rendering/limits/forward_renderer/threaded_render_minimum_instances"); + + _update_shader_quality_settings(); } RenderForwardClustered::~RenderForwardClustered() { diff --git a/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.h b/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.h index 579c8de05e..676f633d33 100644 --- a/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.h +++ b/servers/rendering/renderer_rd/forward_clustered/render_forward_clustered.h @@ -51,6 +51,15 @@ class RenderForwardClustered : public RendererSceneRenderRD { }; enum { + SPEC_CONSTANT_SOFT_SHADOW_SAMPLES = 6, + SPEC_CONSTANT_PENUMBRA_SHADOW_SAMPLES = 7, + SPEC_CONSTANT_DIRECTIONAL_SOFT_SHADOW_SAMPLES = 8, + SPEC_CONSTANT_DIRECTIONAL_PENUMBRA_SHADOW_SAMPLES = 9, + SPEC_CONSTANT_DECAL_FILTER = 10, + SPEC_CONSTANT_PROJECTOR_FILTER = 11, + }; + + enum { SDFGI_MAX_CASCADES = 8, MAX_VOXEL_GI_INSTANCESS = 8, MAX_LIGHTMAPS = 8, @@ -103,12 +112,12 @@ class RenderForwardClustered : public RendererSceneRenderRD { void ensure_specular(); void ensure_voxelgi(); void clear(); - virtual void configure(RID p_color_buffer, RID p_depth_buffer, int p_width, int p_height, RS::ViewportMSAA p_msaa, uint32_t p_view_count); + virtual void configure(RID p_color_buffer, RID p_depth_buffer, RID p_target_buffer, int p_width, int p_height, RS::ViewportMSAA p_msaa, uint32_t p_view_count); ~RenderBufferDataForwardClustered(); }; - virtual RenderBufferData *_create_render_buffer_data(); + virtual RenderBufferData *_create_render_buffer_data() override; void _allocate_normal_roughness_texture(RenderBufferDataForwardClustered *rb); RID render_base_uniform_set; @@ -117,8 +126,8 @@ class RenderForwardClustered : public RendererSceneRenderRD { uint64_t lightmap_texture_array_version = 0xFFFFFFFF; - virtual void _base_uniforms_changed(); - virtual RID _render_buffers_get_normal_texture(RID p_render_buffers); + virtual void _base_uniforms_changed() override; + virtual RID _render_buffers_get_normal_texture(RID p_render_buffers) override; void _update_render_base_uniform_set(); RID _setup_sdfgi_render_pass_uniform_set(RID p_albedo_texture, RID p_emission_texture, RID p_emission_aniso_texture, RID p_geom_facing_texture); @@ -156,8 +165,9 @@ class RenderForwardClustered : public RendererSceneRenderRD { RD::FramebufferFormatID framebuffer_format = 0; uint32_t element_offset = 0; uint32_t barrier = RD::BARRIER_MASK_ALL; + bool use_directional_soft_shadow = false; - RenderListParameters(GeometryInstanceSurfaceDataCache **p_elements, RenderElementInfo *p_element_info, int p_element_count, bool p_reverse_cull, PassMode p_pass_mode, bool p_no_gi, RID p_render_pass_uniform_set, bool p_force_wireframe = false, const Vector2 &p_uv_offset = Vector2(), const Plane &p_lod_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0, uint32_t p_element_offset = 0, uint32_t p_barrier = RD::BARRIER_MASK_ALL) { + RenderListParameters(GeometryInstanceSurfaceDataCache **p_elements, RenderElementInfo *p_element_info, int p_element_count, bool p_reverse_cull, PassMode p_pass_mode, bool p_no_gi, bool p_use_directional_soft_shadows, RID p_render_pass_uniform_set, bool p_force_wireframe = false, const Vector2 &p_uv_offset = Vector2(), const Plane &p_lod_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0, uint32_t p_element_offset = 0, uint32_t p_barrier = RD::BARRIER_MASK_ALL) { elements = p_elements; element_info = p_element_info; element_count = p_element_count; @@ -172,6 +182,7 @@ class RenderForwardClustered : public RendererSceneRenderRD { screen_lod_threshold = p_screen_lod_threshold; element_offset = p_element_offset; barrier = p_barrier; + use_directional_soft_shadow = p_use_directional_soft_shadows; } }; @@ -184,6 +195,7 @@ class RenderForwardClustered : public RendererSceneRenderRD { }; enum { + INSTANCE_DATA_FLAGS_NON_UNIFORM_SCALE = 1 << 5, INSTANCE_DATA_FLAG_USE_GI_BUFFERS = 1 << 6, INSTANCE_DATA_FLAG_USE_SDFGI = 1 << 7, INSTANCE_DATA_FLAG_USE_LIGHTMAP_CAPTURE = 1 << 8, @@ -196,7 +208,6 @@ class RenderForwardClustered : public RendererSceneRenderRD { INSTANCE_DATA_FLAG_MULTIMESH_HAS_CUSTOM_DATA = 1 << 15, INSTANCE_DATA_FLAGS_PARTICLE_TRAIL_SHIFT = 16, INSTANCE_DATA_FLAGS_PARTICLE_TRAIL_MASK = 0xFF, - INSTANCE_DATA_FLAGS_NON_UNIFORM_SCALE = 1 << 24, }; struct SceneState { @@ -220,11 +231,6 @@ class RenderForwardClustered : public RendererSceneRenderRD { float penumbra_shadow_kernel[128]; float soft_shadow_kernel[128]; - uint32_t directional_penumbra_shadow_samples; - uint32_t directional_soft_shadow_samples; - uint32_t penumbra_shadow_samples; - uint32_t soft_shadow_samples; - float ambient_light_color_energy[4]; float ambient_color_sky_mix; @@ -353,7 +359,10 @@ class RenderForwardClustered : public RendererSceneRenderRD { void _setup_lightmaps(const PagedArray<RID> &p_lightmaps, const Transform3D &p_cam_transform); struct RenderElementInfo { - uint32_t repeat : 22; + enum { MAX_REPEATS = (1 << 20) - 1 }; + uint32_t repeat : 20; + uint32_t uses_projector : 1; + uint32_t uses_softshadow : 1; uint32_t uses_lightmap : 1; uint32_t uses_forward_gi : 1; uint32_t lod_index : 8; @@ -371,7 +380,7 @@ class RenderForwardClustered : public RendererSceneRenderRD { uint32_t render_list_thread_threshold = 500; void _update_instance_data_buffer(RenderListType p_render_list); - void _fill_instance_data(RenderListType p_render_list, uint32_t p_offset = 0, int32_t p_max_elements = -1, bool p_update_buffer = true); + void _fill_instance_data(RenderListType p_render_list, int *p_render_info = nullptr, uint32_t p_offset = 0, int32_t p_max_elements = -1, bool p_update_buffer = true); void _fill_render_list(RenderListType p_render_list, const RenderDataRD *p_render_data, PassMode p_pass_mode, bool p_using_sdfgi = false, bool p_using_opaque_gi = false, bool p_append = false); Map<Size2i, RID> sdfgi_framebuffer_size_cache; @@ -402,12 +411,14 @@ class RenderForwardClustered : public RendererSceneRenderRD { union { struct { uint64_t lod_index : 8; - uint64_t surface_index : 10; + uint64_t surface_index : 8; uint64_t geometry_id : 32; - uint64_t material_id_low : 14; + uint64_t material_id_low : 16; - uint64_t material_id_hi : 18; + uint64_t material_id_hi : 16; uint64_t shader_id : 32; + uint64_t uses_softshadow : 1; + uint64_t uses_projector : 1; uint64_t uses_forward_gi : 1; uint64_t uses_lightmap : 1; uint64_t depth_layer : 4; @@ -455,6 +466,8 @@ class RenderForwardClustered : public RendererSceneRenderRD { uint32_t trail_steps = 1; RID mesh_instance; bool can_sdfgi = false; + bool using_projectors = false; + bool using_softshadows = false; //used during setup uint32_t base_flags = 0; Transform3D transform; @@ -551,7 +564,7 @@ class RenderForwardClustered : public RendererSceneRenderRD { } }; - void sort_by_reverse_depth_and_priority(bool p_alpha) { //used for alpha + void sort_by_reverse_depth_and_priority() { //used for alpha SortArray<GeometryInstanceSurfaceDataCache *, SortByReverseDepthAndPriority> sorter; sorter.sort(elements.ptr(), elements.size()); @@ -564,47 +577,51 @@ class RenderForwardClustered : public RendererSceneRenderRD { RenderList render_list[RENDER_LIST_MAX]; + virtual void _update_shader_quality_settings() override; + protected: - virtual void _render_scene(RenderDataRD *p_render_data, const Color &p_default_bg_color); + virtual void _render_scene(RenderDataRD *p_render_data, const Color &p_default_bg_color) override; - virtual void _render_shadow_begin(); - virtual void _render_shadow_append(RID p_framebuffer, const PagedArray<GeometryInstance *> &p_instances, const CameraMatrix &p_projection, const Transform3D &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0, const Rect2i &p_rect = Rect2i(), bool p_flip_y = false, bool p_clear_region = true, bool p_begin = true, bool p_end = true); - virtual void _render_shadow_process(); - virtual void _render_shadow_end(uint32_t p_barrier = RD::BARRIER_MASK_ALL); + virtual void _render_shadow_begin() override; + virtual void _render_shadow_append(RID p_framebuffer, const PagedArray<GeometryInstance *> &p_instances, const CameraMatrix &p_projection, const Transform3D &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0, const Rect2i &p_rect = Rect2i(), bool p_flip_y = false, bool p_clear_region = true, bool p_begin = true, bool p_end = true, RendererScene::RenderInfo *p_render_info = nullptr) override; + virtual void _render_shadow_process() override; + virtual void _render_shadow_end(uint32_t p_barrier = RD::BARRIER_MASK_ALL) override; - virtual void _render_material(const Transform3D &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region); - virtual void _render_uv2(const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region); - virtual void _render_sdfgi(RID p_render_buffers, const Vector3i &p_from, const Vector3i &p_size, const AABB &p_bounds, const PagedArray<GeometryInstance *> &p_instances, const RID &p_albedo_texture, const RID &p_emission_texture, const RID &p_emission_aniso_texture, const RID &p_geom_facing_texture); - virtual void _render_particle_collider_heightfield(RID p_fb, const Transform3D &p_cam_transform, const CameraMatrix &p_cam_projection, const PagedArray<GeometryInstance *> &p_instances); + virtual void _render_material(const Transform3D &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) override; + virtual void _render_uv2(const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) override; + virtual void _render_sdfgi(RID p_render_buffers, const Vector3i &p_from, const Vector3i &p_size, const AABB &p_bounds, const PagedArray<GeometryInstance *> &p_instances, const RID &p_albedo_texture, const RID &p_emission_texture, const RID &p_emission_aniso_texture, const RID &p_geom_facing_texture) override; + virtual void _render_particle_collider_heightfield(RID p_fb, const Transform3D &p_cam_transform, const CameraMatrix &p_cam_projection, const PagedArray<GeometryInstance *> &p_instances) override; public: - virtual GeometryInstance *geometry_instance_create(RID p_base); - virtual void geometry_instance_set_skeleton(GeometryInstance *p_geometry_instance, RID p_skeleton); - virtual void geometry_instance_set_material_override(GeometryInstance *p_geometry_instance, RID p_override); - virtual void geometry_instance_set_surface_materials(GeometryInstance *p_geometry_instance, const Vector<RID> &p_materials); - virtual void geometry_instance_set_mesh_instance(GeometryInstance *p_geometry_instance, RID p_mesh_instance); - virtual void geometry_instance_set_transform(GeometryInstance *p_geometry_instance, const Transform3D &p_transform, const AABB &p_aabb, const AABB &p_transformed_aabb); - virtual void geometry_instance_set_layer_mask(GeometryInstance *p_geometry_instance, uint32_t p_layer_mask); - virtual void geometry_instance_set_lod_bias(GeometryInstance *p_geometry_instance, float p_lod_bias); - virtual void geometry_instance_set_use_baked_light(GeometryInstance *p_geometry_instance, bool p_enable); - virtual void geometry_instance_set_use_dynamic_gi(GeometryInstance *p_geometry_instance, bool p_enable); - virtual void geometry_instance_set_use_lightmap(GeometryInstance *p_geometry_instance, RID p_lightmap_instance, const Rect2 &p_lightmap_uv_scale, int p_lightmap_slice_index); - virtual void geometry_instance_set_lightmap_capture(GeometryInstance *p_geometry_instance, const Color *p_sh9); - virtual void geometry_instance_set_instance_shader_parameters_offset(GeometryInstance *p_geometry_instance, int32_t p_offset); - virtual void geometry_instance_set_cast_double_sided_shadows(GeometryInstance *p_geometry_instance, bool p_enable); - - virtual Transform3D geometry_instance_get_transform(GeometryInstance *p_instance); - virtual AABB geometry_instance_get_aabb(GeometryInstance *p_instance); - - virtual void geometry_instance_free(GeometryInstance *p_geometry_instance); - - virtual uint32_t geometry_instance_get_pair_mask(); - virtual void geometry_instance_pair_light_instances(GeometryInstance *p_geometry_instance, const RID *p_light_instances, uint32_t p_light_instance_count); - virtual void geometry_instance_pair_reflection_probe_instances(GeometryInstance *p_geometry_instance, const RID *p_reflection_probe_instances, uint32_t p_reflection_probe_instance_count); - virtual void geometry_instance_pair_decal_instances(GeometryInstance *p_geometry_instance, const RID *p_decal_instances, uint32_t p_decal_instance_count); - virtual void geometry_instance_pair_voxel_gi_instances(GeometryInstance *p_geometry_instance, const RID *p_voxel_gi_instances, uint32_t p_voxel_gi_instance_count); - - virtual bool free(RID p_rid); + virtual GeometryInstance *geometry_instance_create(RID p_base) override; + virtual void geometry_instance_set_skeleton(GeometryInstance *p_geometry_instance, RID p_skeleton) override; + virtual void geometry_instance_set_material_override(GeometryInstance *p_geometry_instance, RID p_override) override; + virtual void geometry_instance_set_surface_materials(GeometryInstance *p_geometry_instance, const Vector<RID> &p_materials) override; + virtual void geometry_instance_set_mesh_instance(GeometryInstance *p_geometry_instance, RID p_mesh_instance) override; + virtual void geometry_instance_set_transform(GeometryInstance *p_geometry_instance, const Transform3D &p_transform, const AABB &p_aabb, const AABB &p_transformed_aabb) override; + virtual void geometry_instance_set_layer_mask(GeometryInstance *p_geometry_instance, uint32_t p_layer_mask) override; + virtual void geometry_instance_set_lod_bias(GeometryInstance *p_geometry_instance, float p_lod_bias) override; + virtual void geometry_instance_set_use_baked_light(GeometryInstance *p_geometry_instance, bool p_enable) override; + virtual void geometry_instance_set_use_dynamic_gi(GeometryInstance *p_geometry_instance, bool p_enable) override; + virtual void geometry_instance_set_use_lightmap(GeometryInstance *p_geometry_instance, RID p_lightmap_instance, const Rect2 &p_lightmap_uv_scale, int p_lightmap_slice_index) override; + virtual void geometry_instance_set_lightmap_capture(GeometryInstance *p_geometry_instance, const Color *p_sh9) override; + virtual void geometry_instance_set_instance_shader_parameters_offset(GeometryInstance *p_geometry_instance, int32_t p_offset) override; + virtual void geometry_instance_set_cast_double_sided_shadows(GeometryInstance *p_geometry_instance, bool p_enable) override; + + virtual Transform3D geometry_instance_get_transform(GeometryInstance *p_instance) override; + virtual AABB geometry_instance_get_aabb(GeometryInstance *p_instance) override; + + virtual void geometry_instance_free(GeometryInstance *p_geometry_instance) override; + + virtual uint32_t geometry_instance_get_pair_mask() override; + virtual void geometry_instance_pair_light_instances(GeometryInstance *p_geometry_instance, const RID *p_light_instances, uint32_t p_light_instance_count) override; + virtual void geometry_instance_pair_reflection_probe_instances(GeometryInstance *p_geometry_instance, const RID *p_reflection_probe_instances, uint32_t p_reflection_probe_instance_count) override; + virtual void geometry_instance_pair_decal_instances(GeometryInstance *p_geometry_instance, const RID *p_decal_instances, uint32_t p_decal_instance_count) override; + virtual void geometry_instance_pair_voxel_gi_instances(GeometryInstance *p_geometry_instance, const RID *p_voxel_gi_instances, uint32_t p_voxel_gi_instance_count) override; + + virtual void geometry_instance_set_softshadow_projector_pairing(GeometryInstance *p_geometry_instance, bool p_softshadow, bool p_projector) override; + + virtual bool free(RID p_rid) override; RenderForwardClustered(RendererStorageRD *p_storage); ~RenderForwardClustered(); diff --git a/servers/rendering/renderer_rd/forward_clustered/scene_shader_forward_clustered.cpp b/servers/rendering/renderer_rd/forward_clustered/scene_shader_forward_clustered.cpp index f125931df8..a24860996c 100644 --- a/servers/rendering/renderer_rd/forward_clustered/scene_shader_forward_clustered.cpp +++ b/servers/rendering/renderer_rd/forward_clustered/scene_shader_forward_clustered.cpp @@ -131,8 +131,7 @@ void SceneShaderForwardClustered::ShaderData::set_code(const String &p_code) { SceneShaderForwardClustered *shader_singleton = (SceneShaderForwardClustered *)SceneShaderForwardClustered::singleton; Error err = shader_singleton->compiler.compile(RS::SHADER_SPATIAL, code, &actions, path, gen_code); - - ERR_FAIL_COND(err != OK); + ERR_FAIL_COND_MSG(err != OK, "Shader compilation failed."); if (version.is_null()) { version = shader_singleton->shader.version_create(); @@ -287,7 +286,7 @@ void SceneShaderForwardClustered::ShaderData::set_code(const String &p_code) { multisample_state.enable_alpha_to_one = true; } - if (k == SHADER_VERSION_COLOR_PASS || k == SHADER_VERSION_COLOR_PASS_WITH_FORWARD_GI || k == SHADER_VERSION_LIGHTMAP_COLOR_PASS) { + if (k == SHADER_VERSION_COLOR_PASS || k == SHADER_VERSION_LIGHTMAP_COLOR_PASS) { blend_state = blend_state_blend; if (depth_draw == DEPTH_DRAW_OPAQUE) { depth_stencil.enable_depth_write = false; //alpha does not draw depth @@ -305,7 +304,7 @@ void SceneShaderForwardClustered::ShaderData::set_code(const String &p_code) { continue; // do not use this version (will error if using it is attempted) } } else { - if (k == SHADER_VERSION_COLOR_PASS || k == SHADER_VERSION_COLOR_PASS_WITH_FORWARD_GI || k == SHADER_VERSION_LIGHTMAP_COLOR_PASS) { + if (k == SHADER_VERSION_COLOR_PASS || k == SHADER_VERSION_LIGHTMAP_COLOR_PASS) { blend_state = blend_state_opaque; } else if (k == SHADER_VERSION_DEPTH_PASS || k == SHADER_VERSION_DEPTH_PASS_DP) { //none, leave empty @@ -320,13 +319,11 @@ void SceneShaderForwardClustered::ShaderData::set_code(const String &p_code) { } else { //specular write blend_state = blend_state_opaque_specular; - depth_stencil.enable_depth_test = false; - depth_stencil.enable_depth_write = false; } } RID shader_variant = shader_singleton->shader.version_get_shader(version, k); - pipelines[i][j][k].setup(shader_variant, primitive_rd, raster_state, multisample_state, depth_stencil, blend_state, 0); + pipelines[i][j][k].setup(shader_variant, primitive_rd, raster_state, multisample_state, depth_stencil, blend_state, 0, singleton->default_specialization_constants); } } } @@ -410,7 +407,8 @@ RS::ShaderNativeSourceCode SceneShaderForwardClustered::ShaderData::get_native_s return shader_singleton->shader.version_get_native_source_code(version); } -SceneShaderForwardClustered::ShaderData::ShaderData() { +SceneShaderForwardClustered::ShaderData::ShaderData() : + shader_list_element(this) { valid = false; uses_screen_texture = false; } @@ -426,6 +424,7 @@ SceneShaderForwardClustered::ShaderData::~ShaderData() { RendererStorageRD::ShaderData *SceneShaderForwardClustered::_create_shader_func() { ShaderData *shader_data = memnew(ShaderData); + singleton->shader_list.add(&shader_data->shader_list_element); return shader_data; } @@ -437,94 +436,14 @@ void SceneShaderForwardClustered::MaterialData::set_next_pass(RID p_pass) { next_pass = p_pass; } -void SceneShaderForwardClustered::MaterialData::update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) { +bool SceneShaderForwardClustered::MaterialData::update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) { SceneShaderForwardClustered *shader_singleton = (SceneShaderForwardClustered *)SceneShaderForwardClustered::singleton; - if ((uint32_t)ubo_data.size() != shader_data->ubo_size) { - p_uniform_dirty = true; - if (uniform_buffer.is_valid()) { - RD::get_singleton()->free(uniform_buffer); - uniform_buffer = RID(); - } - - ubo_data.resize(shader_data->ubo_size); - if (ubo_data.size()) { - uniform_buffer = RD::get_singleton()->uniform_buffer_create(ubo_data.size()); - memset(ubo_data.ptrw(), 0, ubo_data.size()); //clear - } - - //clear previous uniform set - if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - RD::get_singleton()->free(uniform_set); - uniform_set = RID(); - } - } - - //check whether buffer changed - if (p_uniform_dirty && ubo_data.size()) { - update_uniform_buffer(shader_data->uniforms, shader_data->ubo_offsets.ptr(), p_parameters, ubo_data.ptrw(), ubo_data.size(), false); - RD::get_singleton()->buffer_update(uniform_buffer, 0, ubo_data.size(), ubo_data.ptrw(), RD::BARRIER_MASK_RASTER); - } - - uint32_t tex_uniform_count = shader_data->texture_uniforms.size(); - - if ((uint32_t)texture_cache.size() != tex_uniform_count) { - texture_cache.resize(tex_uniform_count); - p_textures_dirty = true; - - //clear previous uniform set - if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - RD::get_singleton()->free(uniform_set); - uniform_set = RID(); - } - } - - if (p_textures_dirty && tex_uniform_count) { - update_textures(p_parameters, shader_data->default_texture_params, shader_data->texture_uniforms, texture_cache.ptrw(), true); - } - - if (shader_data->ubo_size == 0 && shader_data->texture_uniforms.size() == 0) { - // This material does not require an uniform set, so don't create it. - return; - } - - if (!p_textures_dirty && uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - //no reason to update uniform set, only UBO (or nothing) was needed to update - return; - } - - Vector<RD::Uniform> uniforms; - - { - if (shader_data->ubo_size) { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; - u.binding = 0; - u.ids.push_back(uniform_buffer); - uniforms.push_back(u); - } - - const RID *textures = texture_cache.ptrw(); - for (uint32_t i = 0; i < tex_uniform_count; i++) { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.binding = 1 + i; - u.ids.push_back(textures[i]); - uniforms.push_back(u); - } - } - - uniform_set = RD::get_singleton()->uniform_set_create(uniforms, shader_singleton->shader.version_get_shader(shader_data->version, 0), RenderForwardClustered::MATERIAL_UNIFORM_SET); + return update_parameters_uniform_set(p_parameters, p_uniform_dirty, p_textures_dirty, shader_data->uniforms, shader_data->ubo_offsets.ptr(), shader_data->texture_uniforms, shader_data->default_texture_params, shader_data->ubo_size, uniform_set, shader_singleton->shader.version_get_shader(shader_data->version, 0), RenderForwardClustered::MATERIAL_UNIFORM_SET, RD::BARRIER_MASK_RASTER); } SceneShaderForwardClustered::MaterialData::~MaterialData() { - if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - RD::get_singleton()->free(uniform_set); - } - - if (uniform_buffer.is_valid()) { - RD::get_singleton()->free(uniform_buffer); - } + free_parameters_uniform_set(uniform_set); } RendererStorageRD::MaterialData *SceneShaderForwardClustered::_create_material_func(ShaderData *p_shader) { @@ -546,11 +465,9 @@ SceneShaderForwardClustered::~SceneShaderForwardClustered() { RD::get_singleton()->free(default_vec4_xform_buffer); RD::get_singleton()->free(shadow_sampler); - storage->free(wireframe_material_shader); storage->free(overdraw_material_shader); storage->free(default_shader); - storage->free(wireframe_material); storage->free(overdraw_material); storage->free(default_material); } @@ -567,7 +484,6 @@ void SceneShaderForwardClustered::init(RendererStorageRD *p_storage, const Strin shader_versions.push_back("\n#define MODE_RENDER_DEPTH\n#define MODE_RENDER_MATERIAL\n"); // SHADER_VERSION_DEPTH_PASS_WITH_MATERIAL shader_versions.push_back("\n#define MODE_RENDER_DEPTH\n#define MODE_RENDER_SDF\n"); // SHADER_VERSION_DEPTH_PASS_WITH_SDF shader_versions.push_back(""); // SHADER_VERSION_COLOR_PASS - shader_versions.push_back("\n#define USE_FORWARD_GI\n"); // SHADER_VERSION_COLOR_PASS_WITH_FORWARD_GI shader_versions.push_back("\n#define MODE_MULTIPLE_RENDER_TARGETS\n"); // SHADER_VERSION_COLOR_PASS_WITH_SEPARATE_SPECULAR shader_versions.push_back("\n#define USE_LIGHTMAP\n"); // SHADER_VERSION_LIGHTMAP_COLOR_PASS shader_versions.push_back("\n#define MODE_MULTIPLE_RENDER_TARGETS\n#define USE_LIGHTMAP\n"); // SHADER_VERSION_LIGHTMAP_COLOR_PASS_WITH_SEPARATE_SPECULAR @@ -633,7 +549,6 @@ void SceneShaderForwardClustered::init(RendererStorageRD *p_storage, const Strin actions.renames["SSS_STRENGTH"] = "sss_strength"; actions.renames["SSS_TRANSMITTANCE_COLOR"] = "transmittance_color"; actions.renames["SSS_TRANSMITTANCE_DEPTH"] = "transmittance_depth"; - actions.renames["SSS_TRANSMITTANCE_CURVE"] = "transmittance_curve"; actions.renames["SSS_TRANSMITTANCE_BOOST"] = "transmittance_boost"; actions.renames["BACKLIGHT"] = "backlight"; actions.renames["AO"] = "ao"; @@ -767,7 +682,21 @@ void SceneShaderForwardClustered::init(RendererStorageRD *p_storage, const Strin //default material and shader default_shader = storage->shader_allocate(); storage->shader_initialize(default_shader); - storage->shader_set_code(default_shader, "shader_type spatial; void vertex() { ROUGHNESS = 0.8; } void fragment() { ALBEDO=vec3(0.6); ROUGHNESS=0.8; METALLIC=0.2; } \n"); + storage->shader_set_code(default_shader, R"( +// Default 3D material shader (clustered). + +shader_type spatial; + +void vertex() { + ROUGHNESS = 0.8; +} + +void fragment() { + ALBEDO = vec3(0.6); + ROUGHNESS = 0.8; + METALLIC = 0.2; +} +)"); default_material = storage->material_allocate(); storage->material_initialize(default_material); storage->material_set_shader(default_material, default_shader); @@ -775,22 +704,34 @@ void SceneShaderForwardClustered::init(RendererStorageRD *p_storage, const Strin MaterialData *md = (MaterialData *)storage->material_get_data(default_material, RendererStorageRD::SHADER_TYPE_3D); default_shader_rd = shader.version_get_shader(md->shader_data->version, SHADER_VERSION_COLOR_PASS); default_shader_sdfgi_rd = shader.version_get_shader(md->shader_data->version, SHADER_VERSION_DEPTH_PASS_WITH_SDF); + + default_material_shader_ptr = md->shader_data; + default_material_uniform_set = md->uniform_set; } { overdraw_material_shader = storage->shader_allocate(); storage->shader_initialize(overdraw_material_shader); - storage->shader_set_code(overdraw_material_shader, "shader_type spatial;\nrender_mode blend_add,unshaded;\n void fragment() { ALBEDO=vec3(0.4,0.8,0.8); ALPHA=0.2; }"); + // Use relatively low opacity so that more "layers" of overlapping objects can be distinguished. + storage->shader_set_code(overdraw_material_shader, R"( +// 3D editor Overdraw debug draw mode shader (clustered). + +shader_type spatial; + +render_mode blend_add, unshaded; + +void fragment() { + ALBEDO = vec3(0.4, 0.8, 0.8); + ALPHA = 0.1; +} +)"); overdraw_material = storage->material_allocate(); storage->material_initialize(overdraw_material); storage->material_set_shader(overdraw_material, overdraw_material_shader); - wireframe_material_shader = storage->shader_allocate(); - storage->shader_initialize(wireframe_material_shader); - storage->shader_set_code(wireframe_material_shader, "shader_type spatial;\nrender_mode wireframe,unshaded;\n void fragment() { ALBEDO=vec3(0.0,0.0,0.0); }"); - wireframe_material = storage->material_allocate(); - storage->material_initialize(wireframe_material); - storage->material_set_shader(wireframe_material, wireframe_material_shader); + MaterialData *md = (MaterialData *)storage->material_get_data(overdraw_material, RendererStorageRD::SHADER_TYPE_3D); + overdraw_material_shader_ptr = md->shader_data; + overdraw_material_uniform_set = md->uniform_set; } { @@ -813,3 +754,16 @@ void SceneShaderForwardClustered::init(RendererStorageRD *p_storage, const Strin shadow_sampler = RD::get_singleton()->sampler_create(sampler); } } + +void SceneShaderForwardClustered::set_default_specialization_constants(const Vector<RD::PipelineSpecializationConstant> &p_constants) { + default_specialization_constants = p_constants; + for (SelfList<ShaderData> *E = shader_list.first(); E; E = E->next()) { + for (int i = 0; i < ShaderData::CULL_VARIANT_MAX; i++) { + for (int j = 0; j < RS::PRIMITIVE_MAX; j++) { + for (int k = 0; k < SHADER_VERSION_MAX; k++) { + E->self()->pipelines[i][j][k].update_specialization_constants(default_specialization_constants); + } + } + } + } +} diff --git a/servers/rendering/renderer_rd/forward_clustered/scene_shader_forward_clustered.h b/servers/rendering/renderer_rd/forward_clustered/scene_shader_forward_clustered.h index 8add9f8095..8d75f30a20 100644 --- a/servers/rendering/renderer_rd/forward_clustered/scene_shader_forward_clustered.h +++ b/servers/rendering/renderer_rd/forward_clustered/scene_shader_forward_clustered.h @@ -52,7 +52,6 @@ public: SHADER_VERSION_DEPTH_PASS_WITH_MATERIAL, SHADER_VERSION_DEPTH_PASS_WITH_SDF, SHADER_VERSION_COLOR_PASS, - SHADER_VERSION_COLOR_PASS_WITH_FORWARD_GI, SHADER_VERSION_COLOR_PASS_WITH_SEPARATE_SPECULAR, SHADER_VERSION_LIGHTMAP_COLOR_PASS, SHADER_VERSION_LIGHTMAP_COLOR_PASS_WITH_SEPARATE_SPECULAR, @@ -60,6 +59,13 @@ public: SHADER_VERSION_MAX }; + enum ShaderSpecializations { + SHADER_SPECIALIZATION_FORWARD_GI = 1 << 0, + SHADER_SPECIALIZATION_PROJECTOR = 1 << 1, + SHADER_SPECIALIZATION_SOFT_SHADOWS = 1 << 2, + SHADER_SPECIALIZATION_DIRECTIONAL_SOFT_SHADOWS = 1 << 3, + }; + struct ShaderData : public RendererStorageRD::ShaderData { enum BlendMode { //used internally BLEND_MODE_MIX, @@ -154,10 +160,13 @@ public: virtual Variant get_default_parameter(const StringName &p_parameter) const; virtual RS::ShaderNativeSourceCode get_native_source_code() const; + SelfList<ShaderData> shader_list_element; ShaderData(); virtual ~ShaderData(); }; + SelfList<ShaderData>::List shader_list; + RendererStorageRD::ShaderData *_create_shader_func(); static RendererStorageRD::ShaderData *_create_shader_funcs() { return static_cast<SceneShaderForwardClustered *>(singleton)->_create_shader_func(); @@ -166,17 +175,14 @@ public: struct MaterialData : public RendererStorageRD::MaterialData { uint64_t last_frame; ShaderData *shader_data; - RID uniform_buffer; RID uniform_set; - Vector<RID> texture_cache; - Vector<uint8_t> ubo_data; uint64_t last_pass = 0; uint32_t index = 0; RID next_pass; uint8_t priority; virtual void set_render_priority(int p_priority); virtual void set_next_pass(RID p_pass); - virtual void update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty); + virtual bool update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty); virtual ~MaterialData(); }; @@ -192,8 +198,6 @@ public: RID default_material; RID overdraw_material_shader; RID overdraw_material; - RID wireframe_material_shader; - RID wireframe_material; RID default_shader_rd; RID default_shader_sdfgi_rd; @@ -202,10 +206,18 @@ public: RID shadow_sampler; + RID default_material_uniform_set; + ShaderData *default_material_shader_ptr = nullptr; + + RID overdraw_material_uniform_set; + ShaderData *overdraw_material_shader_ptr = nullptr; + + Vector<RD::PipelineSpecializationConstant> default_specialization_constants; SceneShaderForwardClustered(); ~SceneShaderForwardClustered(); void init(RendererStorageRD *p_storage, const String p_defines); + void set_default_specialization_constants(const Vector<RD::PipelineSpecializationConstant> &p_constants); }; } // namespace RendererSceneRenderImplementation diff --git a/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.cpp b/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.cpp index ae09d215ff..276a44bc27 100644 --- a/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.cpp +++ b/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.cpp @@ -35,6 +35,34 @@ using namespace RendererSceneRenderImplementation; +RenderForwardMobile::ForwardID RenderForwardMobile::_allocate_forward_id(ForwardIDType p_type) { + int32_t index = -1; + for (uint32_t i = 0; i < forward_id_allocators[p_type].allocations.size(); i++) { + if (forward_id_allocators[p_type].allocations[i] == false) { + index = i; + break; + } + } + + if (index == -1) { + index = forward_id_allocators[p_type].allocations.size(); + forward_id_allocators[p_type].allocations.push_back(true); + forward_id_allocators[p_type].map.push_back(0xFF); + } else { + forward_id_allocators[p_type].allocations[index] = true; + } + + return index; +} +void RenderForwardMobile::_free_forward_id(ForwardIDType p_type, ForwardID p_id) { + ERR_FAIL_INDEX(p_id, (ForwardID)forward_id_allocators[p_type].allocations.size()); + forward_id_allocators[p_type].allocations[p_id] = false; +} + +void RenderForwardMobile::_map_forward_id(ForwardIDType p_type, ForwardID p_id, uint32_t p_index) { + forward_id_allocators[p_type].map[p_id] = p_index; +} + /* Render buffer */ void RenderForwardMobile::RenderBufferDataForwardMobile::clear() { @@ -50,37 +78,77 @@ void RenderForwardMobile::RenderBufferDataForwardMobile::clear() { color = RID(); depth = RID(); - color_fb = RID(); + for (int i = 0; i < FB_CONFIG_MAX; i++) { + color_fbs[i] = RID(); + } } -void RenderForwardMobile::RenderBufferDataForwardMobile::configure(RID p_color_buffer, RID p_depth_buffer, int p_width, int p_height, RS::ViewportMSAA p_msaa, uint32_t p_view_count) { +void RenderForwardMobile::RenderBufferDataForwardMobile::configure(RID p_color_buffer, RID p_depth_buffer, RID p_target_buffer, int p_width, int p_height, RS::ViewportMSAA p_msaa, uint32_t p_view_count) { clear(); msaa = p_msaa; + Size2i target_size = RD::get_singleton()->texture_size(p_target_buffer); + width = p_width; height = p_height; + bool is_scaled = (target_size.width != p_width) || (target_size.height != p_height); view_count = p_view_count; color = p_color_buffer; depth = p_depth_buffer; - // re-introduce setting up msaa? For now we ignore this... + // We are creating 4 configurations here for our framebuffers. if (p_msaa == RS::VIEWPORT_MSAA_DISABLED) { Vector<RID> fb; - fb.push_back(p_color_buffer); - fb.push_back(depth); + fb.push_back(p_color_buffer); // 0 - color buffer + fb.push_back(depth); // 1 - depth buffer + + // Now define our subpasses + Vector<RD::FramebufferPass> passes; + RD::FramebufferPass pass; - color_fb = RD::get_singleton()->framebuffer_create(fb, RenderingDevice::INVALID_ID, view_count); + // re-using the same attachments + pass.color_attachments.push_back(0); + pass.depth_attachment = 1; + + // - opaque pass + passes.push_back(pass); + color_fbs[FB_CONFIG_ONE_PASS] = RD::get_singleton()->framebuffer_create_multipass(fb, passes, RenderingDevice::INVALID_ID, view_count); + + // - add sky pass + passes.push_back(pass); + color_fbs[FB_CONFIG_TWO_SUBPASSES] = RD::get_singleton()->framebuffer_create_multipass(fb, passes, RenderingDevice::INVALID_ID, view_count); + + // - add alpha pass + passes.push_back(pass); + color_fbs[FB_CONFIG_THREE_SUBPASSES] = RD::get_singleton()->framebuffer_create_multipass(fb, passes, RenderingDevice::INVALID_ID, view_count); + + if (!is_scaled) { + // - add blit to 2D pass + fb.push_back(p_target_buffer); // 2 - target buffer + + RD::FramebufferPass blit_pass; + blit_pass.color_attachments.push_back(2); + blit_pass.input_attachments.push_back(0); + passes.push_back(blit_pass); + + color_fbs[FB_CONFIG_FOUR_SUBPASSES] = RD::get_singleton()->framebuffer_create_multipass(fb, passes, RenderingDevice::INVALID_ID, view_count); + } else { + // can't do our blit pass if resolutions don't match + color_fbs[FB_CONFIG_FOUR_SUBPASSES] = RID(); + } } else { + RD::DataFormat color_format = RenderForwardMobile::singleton->_render_buffers_get_color_format(); + RD::TextureFormat tf; if (view_count > 1) { tf.texture_type = RD::TEXTURE_TYPE_2D_ARRAY; } else { tf.texture_type = RD::TEXTURE_TYPE_2D; } - tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + tf.format = color_format; tf.width = p_width; tf.height = p_height; tf.array_layers = view_count; // create a layer for every view @@ -106,14 +174,87 @@ void RenderForwardMobile::RenderBufferDataForwardMobile::configure(RID p_color_b { Vector<RID> fb; - fb.push_back(color_msaa); - fb.push_back(depth_msaa); + fb.push_back(color_msaa); // 0 - msaa color buffer + fb.push_back(depth_msaa); // 1 - msaa depth buffer + + // Now define our subpasses + Vector<RD::FramebufferPass> passes; + RD::FramebufferPass pass; + + // re-using the same attachments + pass.color_attachments.push_back(0); + pass.depth_attachment = 1; + + // - opaque pass + passes.push_back(pass); + + // - add sky pass + fb.push_back(color); // 2 - color buffer + passes.push_back(pass); // without resolve for our 3 + 4 subpass config + { + // but with resolve for our 2 subpass config + Vector<RD::FramebufferPass> two_passes; + two_passes.push_back(pass); // opaque subpass without resolve + pass.resolve_attachments.push_back(2); + two_passes.push_back(pass); // sky subpass with resolve + + color_fbs[FB_CONFIG_TWO_SUBPASSES] = RD::get_singleton()->framebuffer_create_multipass(fb, two_passes, RenderingDevice::INVALID_ID, view_count); + } - color_fb = RD::get_singleton()->framebuffer_create(fb, RenderingDevice::INVALID_ID, view_count); + // - add alpha pass (with resolve, we just added that above) + passes.push_back(pass); + color_fbs[FB_CONFIG_THREE_SUBPASSES] = RD::get_singleton()->framebuffer_create_multipass(fb, passes, RenderingDevice::INVALID_ID, view_count); + + { + // we also need our one pass with resolve + Vector<RD::FramebufferPass> one_pass_with_resolve; + one_pass_with_resolve.push_back(pass); // note our pass configuration already has resolve.. + color_fbs[FB_CONFIG_ONE_PASS] = RD::get_singleton()->framebuffer_create_multipass(fb, one_pass_with_resolve, RenderingDevice::INVALID_ID, view_count); + } + + if (!is_scaled) { + // - add blit to 2D pass + fb.push_back(p_target_buffer); // 3 - target buffer + RD::FramebufferPass blit_pass; + blit_pass.color_attachments.push_back(3); + blit_pass.input_attachments.push_back(2); + passes.push_back(blit_pass); + + color_fbs[FB_CONFIG_FOUR_SUBPASSES] = RD::get_singleton()->framebuffer_create_multipass(fb, passes, RenderingDevice::INVALID_ID, view_count); + } else { + // can't do our blit pass if resolutions don't match + color_fbs[FB_CONFIG_FOUR_SUBPASSES] = RID(); + } } } } +RID RenderForwardMobile::reflection_probe_create_framebuffer(RID p_color, RID p_depth) { + // Our attachments + Vector<RID> fb; + fb.push_back(p_color); // 0 + fb.push_back(p_depth); // 1 + + // Now define our subpasses + Vector<RD::FramebufferPass> passes; + RD::FramebufferPass pass; + + // re-using the same attachments + pass.color_attachments.push_back(0); + pass.depth_attachment = 1; + + // - opaque pass + passes.push_back(pass); + + // - sky pass + passes.push_back(pass); + + // - alpha pass + passes.push_back(pass); + + return RD::get_singleton()->framebuffer_create_multipass(fb, passes); +} + RenderForwardMobile::RenderBufferDataForwardMobile::~RenderBufferDataForwardMobile() { clear(); } @@ -131,6 +272,23 @@ bool RenderForwardMobile::free(RID p_rid) { /* Render functions */ +float RenderForwardMobile::_render_buffers_get_luminance_multiplier() { + // On mobile renderer we need to multiply source colors by 2 due to using a UNORM buffer + // and multiplying by the output color during 3D rendering by 0.5 + return 2.0; +} + +RD::DataFormat RenderForwardMobile::_render_buffers_get_color_format() { + // Using 32bit buffers enables AFBC on mobile devices which should have a definite performance improvement (MALI G710 and newer support this on 64bit RTs) + return RD::DATA_FORMAT_A2B10G10R10_UNORM_PACK32; +} + +bool RenderForwardMobile::_render_buffers_can_be_storage() { + // Using 32bit buffers enables AFBC on mobile devices which should have a definite performance improvement (MALI G710 and newer support this on 64bit RTs) + // Doesn't support storage + return false; +} + RID RenderForwardMobile::_setup_render_pass_uniform_set(RenderListType p_render_list, const RenderDataRD *p_render_data, RID p_radiance_texture, bool p_use_directional_shadow_atlas, int p_index) { //there should always be enough uniform buffers for render passes, otherwise bugs ERR_FAIL_INDEX_V(p_index, (int)scene_state.uniform_buffers.size(), RID()); @@ -263,7 +421,8 @@ RID RenderForwardMobile::_setup_render_pass_uniform_set(RenderListType p_render_ RD::Uniform u; u.binding = 9; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - RID texture = (false && rb && rb->depth.is_valid()) ? rb->depth : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE); + RID dbt = rb ? render_buffers_get_back_depth_texture(p_render_data->render_buffers) : RID(); + RID texture = (dbt.is_valid()) ? dbt : storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE); u.ids.push_back(texture); uniforms.push_back(u); } @@ -326,29 +485,69 @@ void RenderForwardMobile::_render_scene(RenderDataRD *p_render_data, const Color scene_state.ubo.viewport_size[1] = vp_he.y; scene_state.ubo.directional_light_count = 0; + // We can only use our full subpass approach if we're: + // - not reading from SCREEN_TEXTURE/DEPTH_TEXTURE + // - not using ssr/sss (currently not supported) + // - not using glow or other post effects (can't do 4th subpass) + // - rendering to a half sized render buffer (can't do 4th subpass) + // We'll need to restrict how far we're going with subpasses based on this. + Size2i screen_size; - RID opaque_framebuffer; - RID alpha_framebuffer; + RID framebuffer; bool reverse_cull = false; + bool using_subpass_transparent = true; + bool using_subpass_post_process = true; - // I don't think we support either of these in our mobile renderer so probably should phase them out - bool using_ssr = false; - bool using_sss = false; + bool using_ssr = false; // I don't think we support this in our mobile renderer so probably should phase it out + bool using_sss = false; // I don't think we support this in our mobile renderer so probably should phase it out + + // fill our render lists early so we can find out if we use various features + _fill_render_list(RENDER_LIST_OPAQUE, p_render_data, PASS_MODE_COLOR); + render_list[RENDER_LIST_OPAQUE].sort_by_key(); + render_list[RENDER_LIST_ALPHA].sort_by_reverse_depth_and_priority(); + _fill_element_info(RENDER_LIST_OPAQUE); + _fill_element_info(RENDER_LIST_ALPHA); + + if (p_render_data->render_info) { + p_render_data->render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_VISIBLE][RS::VIEWPORT_RENDER_INFO_DRAW_CALLS_IN_FRAME] = p_render_data->instances->size(); + p_render_data->render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_VISIBLE][RS::VIEWPORT_RENDER_INFO_OBJECTS_IN_FRAME] = p_render_data->instances->size(); + } if (render_buffer) { // setup rendering to render buffer screen_size.x = render_buffer->width; screen_size.y = render_buffer->height; - opaque_framebuffer = render_buffer->color_fb; - alpha_framebuffer = opaque_framebuffer; + if (render_buffer->color_fbs[FB_CONFIG_FOUR_SUBPASSES].is_null()) { + // can't do blit subpass + using_subpass_post_process = false; + } else if (env && (env->glow_enabled || env->auto_exposure || camera_effects_uses_dof(p_render_data->camera_effects))) { + // can't do blit subpass + using_subpass_post_process = false; + } + + if (using_ssr || using_sss || scene_state.used_screen_texture || scene_state.used_depth_texture) { + // can't use our last two subpasses + using_subpass_transparent = false; + using_subpass_post_process = false; + } + + if (using_subpass_post_process) { + // all as subpasses + framebuffer = render_buffer->color_fbs[FB_CONFIG_FOUR_SUBPASSES]; + } else if (using_subpass_transparent) { + // our tonemap pass is separate + framebuffer = render_buffer->color_fbs[FB_CONFIG_THREE_SUBPASSES]; + } else { + // only opaque and sky as subpasses + framebuffer = render_buffer->color_fbs[FB_CONFIG_TWO_SUBPASSES]; + } } else if (p_render_data->reflection_probe.is_valid()) { uint32_t resolution = reflection_probe_instance_get_resolution(p_render_data->reflection_probe); screen_size.x = resolution; screen_size.y = resolution; - opaque_framebuffer = reflection_probe_instance_get_framebuffer(p_render_data->reflection_probe, p_render_data->reflection_probe_pass); - alpha_framebuffer = opaque_framebuffer; + framebuffer = reflection_probe_instance_get_framebuffer(p_render_data->reflection_probe, p_render_data->reflection_probe_pass); if (storage->reflection_probe_is_interior(reflection_probe_instance_get_probe(p_render_data->reflection_probe))) { p_render_data->environment = RID(); //no environment on interiors @@ -356,6 +555,8 @@ void RenderForwardMobile::_render_scene(RenderDataRD *p_render_data, const Color } reverse_cull = true; + using_subpass_transparent = true; // we ignore our screen/depth texture here + using_subpass_post_process = false; // not applicable at all for reflection probes. } else { ERR_FAIL(); //bug? } @@ -367,17 +568,7 @@ void RenderForwardMobile::_render_scene(RenderDataRD *p_render_data, const Color _update_render_base_uniform_set(); //may have changed due to the above (light buffer enlarged, as an example) - _fill_render_list(RENDER_LIST_OPAQUE, p_render_data, PASS_MODE_COLOR); - render_list[RENDER_LIST_OPAQUE].sort_by_key(); - render_list[RENDER_LIST_ALPHA].sort_by_depth(); - - // we no longer use this... - _fill_instance_data(RENDER_LIST_OPAQUE); - _fill_instance_data(RENDER_LIST_ALPHA); - - RD::get_singleton()->draw_command_end_label(); - - // note, no depth prepass here! + RD::get_singleton()->draw_command_end_label(); // Render Setup // setup environment RID radiance_texture; @@ -446,103 +637,234 @@ void RenderForwardMobile::_render_scene(RenderDataRD *p_render_data, const Color RID sky_rid = env->sky; if (sky_rid.is_valid()) { - sky.update(env, projection, p_render_data->cam_transform, time); + sky.update(env, projection, p_render_data->cam_transform, time, _render_buffers_get_luminance_multiplier()); radiance_texture = sky.sky_get_radiance_texture_rd(sky_rid); } else { // do not try to draw sky if invalid draw_sky = false; } - RD::get_singleton()->draw_command_end_label(); + RD::get_singleton()->draw_command_end_label(); // Setup Sky } } else { clear_color = p_default_bg_color; } - // opaque pass + // update sky buffers (if required) + if (draw_sky || draw_sky_fog_only) { + // !BAS! @TODO See if we can limit doing some things double and maybe even move this into _pre_opaque_render + // and change Forward Clustered in the same way as we have here (but without using subpasses) + RENDER_TIMESTAMP("Setup Sky resolution buffers"); - // !BAS! Look into this, seems most of the code in here related to clustered only, may want to move this code into ForwardClustered/RenderForwardMobile before calling it from here - // does trigger shadow map rendering so kinda important - _pre_opaque_render(p_render_data, false, false, RID(), RID()); + RD::get_singleton()->draw_command_begin_label("Setup Sky resolution buffers"); - RD::get_singleton()->draw_command_begin_label("Render Opaque Pass"); + if (p_render_data->reflection_probe.is_valid()) { + CameraMatrix correction; + correction.set_depth_correction(true); + CameraMatrix projection = correction * p_render_data->cam_projection; + sky.update_res_buffers(env, 1, &projection, p_render_data->cam_transform, time); + } else { + sky.update_res_buffers(env, p_render_data->view_count, p_render_data->view_projection, p_render_data->cam_transform, time); + } - scene_state.ubo.directional_light_count = p_render_data->directional_light_count; + RD::get_singleton()->draw_command_end_label(); // Setup Sky resolution buffers + } - _setup_environment(p_render_data, p_render_data->reflection_probe.is_valid(), screen_size, !p_render_data->reflection_probe.is_valid(), p_default_bg_color, p_render_data->render_buffers.is_valid()); + _pre_opaque_render(p_render_data, false, false, RID(), RID()); - RENDER_TIMESTAMP("Render Opaque Pass"); + uint32_t spec_constant_base_flags = 0; - RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_OPAQUE, p_render_data, radiance_texture, true); + { + //figure out spec constants - bool can_continue_color = !scene_state.used_screen_texture && !using_ssr && !using_sss; - bool can_continue_depth = !scene_state.used_depth_texture && !using_ssr && !using_sss; + if (p_render_data->directional_light_count > 0) { + if (p_render_data->directional_light_soft_shadows) { + spec_constant_base_flags |= 1 << SPEC_CONSTANT_USING_DIRECTIONAL_SOFT_SHADOWS; + } + } else { + spec_constant_base_flags |= 1 << SPEC_CONSTANT_DISABLE_DIRECTIONAL_LIGHTS; + } + if (!is_environment(p_render_data->environment) || environment_is_fog_enabled(p_render_data->environment)) { + spec_constant_base_flags |= 1 << SPEC_CONSTANT_DISABLE_FOG; + } + } { - bool will_continue_color = (can_continue_color || draw_sky || draw_sky_fog_only); - bool will_continue_depth = (can_continue_depth || draw_sky || draw_sky_fog_only); + if (render_buffer) { + RD::get_singleton()->draw_command_begin_label("Render 3D Pass"); + } else { + RD::get_singleton()->draw_command_begin_label("Render Reflection Probe Pass"); + } - // regular forward for now - Vector<Color> c; - c.push_back(clear_color.to_linear()); + // opaque pass - RenderListParameters render_list_params(render_list[RENDER_LIST_OPAQUE].elements.ptr(), render_list[RENDER_LIST_OPAQUE].element_info.ptr(), render_list[RENDER_LIST_OPAQUE].elements.size(), reverse_cull, PASS_MODE_COLOR, rp_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), p_render_data->lod_camera_plane, p_render_data->lod_distance_multiplier, p_render_data->screen_lod_threshold, p_render_data->view_count); - _render_list_with_threads(&render_list_params, opaque_framebuffer, keep_color ? RD::INITIAL_ACTION_KEEP : RD::INITIAL_ACTION_CLEAR, will_continue_color ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, will_continue_depth ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, c, 1.0, 0); - } + RD::get_singleton()->draw_command_begin_label("Render Opaque Subpass"); - RD::get_singleton()->draw_command_end_label(); + scene_state.ubo.directional_light_count = p_render_data->directional_light_count; - if (draw_sky || draw_sky_fog_only) { - RENDER_TIMESTAMP("Render Sky"); + _setup_environment(p_render_data, p_render_data->reflection_probe.is_valid(), screen_size, !p_render_data->reflection_probe.is_valid(), p_default_bg_color, p_render_data->render_buffers.is_valid()); - RD::get_singleton()->draw_command_begin_label("Draw Sky"); + if (using_subpass_transparent && using_subpass_post_process) { + RENDER_TIMESTAMP("Render Opaque + Transparent + Tonemap"); + } else if (using_subpass_transparent) { + RENDER_TIMESTAMP("Render Opaque + Transparent"); + } else { + RENDER_TIMESTAMP("Render Opaque"); + } - if (p_render_data->reflection_probe.is_valid()) { - CameraMatrix correction; - correction.set_depth_correction(true); - CameraMatrix projection = correction * p_render_data->cam_projection; - sky.draw(env, can_continue_color, can_continue_depth, opaque_framebuffer, 1, &projection, p_render_data->cam_transform, time); + RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_OPAQUE, p_render_data, radiance_texture, true); + + bool can_continue_color = !using_subpass_transparent && !scene_state.used_screen_texture && !using_ssr && !using_sss; + bool can_continue_depth = !using_subpass_transparent && !scene_state.used_depth_texture && !using_ssr && !using_sss; + + { + // regular forward for now + Vector<Color> c; + c.push_back(clear_color.to_linear()); // our render buffer + if (render_buffer) { + if (render_buffer->msaa != RS::VIEWPORT_MSAA_DISABLED) { + c.push_back(clear_color.to_linear()); // our resolve buffer + } + if (using_subpass_post_process) { + c.push_back(Color()); // our 2D buffer we're copying into + } + } + + RD::FramebufferFormatID fb_format = RD::get_singleton()->framebuffer_get_format(framebuffer); + RenderListParameters render_list_params(render_list[RENDER_LIST_OPAQUE].elements.ptr(), render_list[RENDER_LIST_OPAQUE].element_info.ptr(), render_list[RENDER_LIST_OPAQUE].elements.size(), reverse_cull, PASS_MODE_COLOR, rp_uniform_set, spec_constant_base_flags, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), p_render_data->lod_camera_plane, p_render_data->lod_distance_multiplier, p_render_data->screen_lod_threshold, p_render_data->view_count); + render_list_params.framebuffer_format = fb_format; + if ((uint32_t)render_list_params.element_count > render_list_thread_threshold && false) { + // secondary command buffers need more testing at this time + //multi threaded + thread_draw_lists.resize(RendererThreadPool::singleton->thread_work_pool.get_thread_count()); + RD::get_singleton()->draw_list_begin_split(framebuffer, thread_draw_lists.size(), thread_draw_lists.ptr(), keep_color ? RD::INITIAL_ACTION_KEEP : RD::INITIAL_ACTION_CLEAR, can_continue_color ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, can_continue_depth ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, c, 1.0, 0); + RendererThreadPool::singleton->thread_work_pool.do_work(thread_draw_lists.size(), this, &RenderForwardMobile::_render_list_thread_function, &render_list_params); + } else { + //single threaded + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(framebuffer, keep_color ? RD::INITIAL_ACTION_KEEP : RD::INITIAL_ACTION_CLEAR, can_continue_color ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, can_continue_depth ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, c, 1.0, 0); + _render_list(draw_list, fb_format, &render_list_params, 0, render_list_params.element_count); + } + } + + RD::get_singleton()->draw_command_end_label(); //Render Opaque Subpass + + if (draw_sky || draw_sky_fog_only) { + RD::get_singleton()->draw_command_begin_label("Draw Sky Subpass"); + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_switch_to_next_pass(); + + if (p_render_data->reflection_probe.is_valid()) { + CameraMatrix correction; + correction.set_depth_correction(true); + CameraMatrix projection = correction * p_render_data->cam_projection; + sky.draw(draw_list, env, framebuffer, 1, &projection, p_render_data->cam_transform, time, _render_buffers_get_luminance_multiplier()); + } else { + sky.draw(draw_list, env, framebuffer, p_render_data->view_count, p_render_data->view_projection, p_render_data->cam_transform, time, _render_buffers_get_luminance_multiplier()); + } + + RD::get_singleton()->draw_command_end_label(); // Draw Sky Subpass + + // note, if MSAA is used in 2-subpass approach we should get an automatic resolve here } else { - sky.draw(env, can_continue_color, can_continue_depth, opaque_framebuffer, p_render_data->view_count, p_render_data->view_projection, p_render_data->cam_transform, time); + // switch to subpass but we do nothing here so basically we skip (though this should trigger resolve with 2-subpass MSAA). + RD::get_singleton()->draw_list_switch_to_next_pass(); } - RD::get_singleton()->draw_command_end_label(); - } - if (render_buffer && !can_continue_color && render_buffer->msaa != RS::VIEWPORT_MSAA_DISABLED) { - RD::get_singleton()->texture_resolve_multisample(render_buffer->color_msaa, render_buffer->color); - /* - if (using_separate_specular) { - RD::get_singleton()->texture_resolve_multisample(render_buffer->specular_msaa, render_buffer->specular); + if (!using_subpass_transparent) { + // We're done with our subpasses so end our container pass + RD::get_singleton()->draw_list_end(RD::BARRIER_MASK_ALL); + + RD::get_singleton()->draw_command_end_label(); // Render 3D Pass / Render Reflection Probe Pass } - */ - } - if (render_buffer && !can_continue_depth && render_buffer->msaa != RS::VIEWPORT_MSAA_DISABLED) { - RD::get_singleton()->texture_resolve_multisample(render_buffer->depth_msaa, render_buffer->depth); - } + if (scene_state.used_screen_texture) { + // Copy screen texture to backbuffer so we can read from it + _render_buffers_copy_screen_texture(p_render_data); + } - // transparent pass - RENDER_TIMESTAMP("Render Transparent Pass"); + if (scene_state.used_depth_texture) { + // Copy depth texture to backbuffer so we can read from it + _render_buffers_copy_depth_texture(p_render_data); + } - RD::get_singleton()->draw_command_begin_label("Render Transparent Pass"); + // transparent pass - rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_ALPHA, p_render_data, radiance_texture, true); + RD::get_singleton()->draw_command_begin_label("Render Transparent Subpass"); - _setup_environment(p_render_data, p_render_data->reflection_probe.is_valid(), screen_size, !p_render_data->reflection_probe.is_valid(), p_default_bg_color, false); + rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_ALPHA, p_render_data, radiance_texture, true); - { - RenderListParameters render_list_params(render_list[RENDER_LIST_ALPHA].elements.ptr(), render_list[RENDER_LIST_ALPHA].element_info.ptr(), render_list[RENDER_LIST_ALPHA].elements.size(), reverse_cull, PASS_MODE_COLOR, rp_uniform_set, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), p_render_data->lod_camera_plane, p_render_data->lod_distance_multiplier, p_render_data->screen_lod_threshold, p_render_data->view_count); - _render_list_with_threads(&render_list_params, alpha_framebuffer, can_continue_color ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, can_continue_depth ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ); + if (using_subpass_transparent) { + RD::FramebufferFormatID fb_format = RD::get_singleton()->framebuffer_get_format(framebuffer); + RenderListParameters render_list_params(render_list[RENDER_LIST_ALPHA].elements.ptr(), render_list[RENDER_LIST_ALPHA].element_info.ptr(), render_list[RENDER_LIST_ALPHA].elements.size(), reverse_cull, PASS_MODE_COLOR, rp_uniform_set, spec_constant_base_flags, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), p_render_data->lod_camera_plane, p_render_data->lod_distance_multiplier, p_render_data->screen_lod_threshold, p_render_data->view_count); + render_list_params.framebuffer_format = fb_format; + if ((uint32_t)render_list_params.element_count > render_list_thread_threshold && false) { + // secondary command buffers need more testing at this time + //multi threaded + thread_draw_lists.resize(RendererThreadPool::singleton->thread_work_pool.get_thread_count()); + RD::get_singleton()->draw_list_switch_to_next_pass_split(thread_draw_lists.size(), thread_draw_lists.ptr()); + render_list_params.subpass = RD::get_singleton()->draw_list_get_current_pass(); + RendererThreadPool::singleton->thread_work_pool.do_work(thread_draw_lists.size(), this, &RenderForwardMobile::_render_list_thread_function, &render_list_params); + } else { + //single threaded + RD::DrawListID draw_list = RD::get_singleton()->draw_list_switch_to_next_pass(); + render_list_params.subpass = RD::get_singleton()->draw_list_get_current_pass(); + _render_list(draw_list, fb_format, &render_list_params, 0, render_list_params.element_count); + } + + RD::get_singleton()->draw_command_end_label(); // Render Transparent Subpass + + // note if we are using MSAA we should get an automatic resolve through our subpass configuration. + + // blit to tonemap + if (render_buffer && using_subpass_post_process) { + _post_process_subpass(render_buffer->color, framebuffer, p_render_data); + } + + RD::get_singleton()->draw_command_end_label(); // Render 3D Pass / Render Reflection Probe Pass + + RD::get_singleton()->draw_list_end(RD::BARRIER_MASK_ALL); + } else { + RENDER_TIMESTAMP("Render Transparent"); + + framebuffer = render_buffer->color_fbs[FB_CONFIG_ONE_PASS]; + + // this may be needed if we re-introduced steps that change info, not sure which do so in the previous implementation + // _setup_environment(p_render_data, p_render_data->reflection_probe.is_valid(), screen_size, !p_render_data->reflection_probe.is_valid(), p_default_bg_color, false); + + RD::FramebufferFormatID fb_format = RD::get_singleton()->framebuffer_get_format(framebuffer); + RenderListParameters render_list_params(render_list[RENDER_LIST_ALPHA].elements.ptr(), render_list[RENDER_LIST_ALPHA].element_info.ptr(), render_list[RENDER_LIST_ALPHA].elements.size(), reverse_cull, PASS_MODE_COLOR, rp_uniform_set, spec_constant_base_flags, get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_WIREFRAME, Vector2(), p_render_data->lod_camera_plane, p_render_data->lod_distance_multiplier, p_render_data->screen_lod_threshold, p_render_data->view_count); + render_list_params.framebuffer_format = fb_format; + if ((uint32_t)render_list_params.element_count > render_list_thread_threshold && false) { + // secondary command buffers need more testing at this time + //multi threaded + thread_draw_lists.resize(RendererThreadPool::singleton->thread_work_pool.get_thread_count()); + RD::get_singleton()->draw_list_begin_split(framebuffer, thread_draw_lists.size(), thread_draw_lists.ptr(), can_continue_color ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, can_continue_depth ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ); + RendererThreadPool::singleton->thread_work_pool.do_work(thread_draw_lists.size(), this, &RenderForwardMobile::_render_list_thread_function, &render_list_params); + RD::get_singleton()->draw_list_end(RD::BARRIER_MASK_ALL); + } else { + //single threaded + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(framebuffer, can_continue_color ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, can_continue_depth ? RD::INITIAL_ACTION_CONTINUE : RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ); + _render_list(draw_list, fb_format, &render_list_params, 0, render_list_params.element_count); + RD::get_singleton()->draw_list_end(RD::BARRIER_MASK_ALL); + } + + RD::get_singleton()->draw_command_end_label(); // Render Transparent Subpass + } } - RD::get_singleton()->draw_command_end_label(); + if (render_buffer && !using_subpass_post_process) { + RD::get_singleton()->draw_command_begin_label("Post process pass"); - RD::get_singleton()->draw_command_begin_label("Resolve"); + // If we need extra effects we do this in its own pass + RENDER_TIMESTAMP("Tonemap"); - if (render_buffer && render_buffer->msaa != RS::VIEWPORT_MSAA_DISABLED) { - RD::get_singleton()->texture_resolve_multisample(render_buffer->color_msaa, render_buffer->color); + _render_buffers_post_process_and_tonemap(p_render_data); + + RD::get_singleton()->draw_command_end_label(); // Post process pass } - RD::get_singleton()->draw_command_end_label(); + if (render_buffer) { + _disable_clear_request(p_render_data); + } } /* these are being called from RendererSceneRenderRD::_pre_opaque_render */ @@ -555,11 +877,15 @@ void RenderForwardMobile::_render_shadow_begin() { render_list[RENDER_LIST_SECONDARY].clear(); } -void RenderForwardMobile::_render_shadow_append(RID p_framebuffer, const PagedArray<GeometryInstance *> &p_instances, const CameraMatrix &p_projection, const Transform3D &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane, float p_lod_distance_multiplier, float p_screen_lod_threshold, const Rect2i &p_rect, bool p_flip_y, bool p_clear_region, bool p_begin, bool p_end) { +void RenderForwardMobile::_render_shadow_append(RID p_framebuffer, const PagedArray<GeometryInstance *> &p_instances, const CameraMatrix &p_projection, const Transform3D &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane, float p_lod_distance_multiplier, float p_screen_lod_threshold, const Rect2i &p_rect, bool p_flip_y, bool p_clear_region, bool p_begin, bool p_end, RendererScene::RenderInfo *p_render_info) { uint32_t shadow_pass_index = scene_state.shadow_passes.size(); SceneState::ShadowPass shadow_pass; + if (p_render_info) { + p_render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_SHADOW][RS::VIEWPORT_RENDER_INFO_DRAW_CALLS_IN_FRAME] = p_instances.size(); + p_render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_SHADOW][RS::VIEWPORT_RENDER_INFO_OBJECTS_IN_FRAME] = p_instances.size(); + } RenderDataRD render_data; render_data.cam_projection = p_projection; render_data.cam_transform = p_transform; @@ -567,6 +893,7 @@ void RenderForwardMobile::_render_shadow_append(RID p_framebuffer, const PagedAr render_data.z_near = 0.0; render_data.z_far = p_zfar; render_data.instances = &p_instances; + render_data.render_info = p_render_info; render_data.lod_camera_plane = p_camera_plane; render_data.lod_distance_multiplier = p_lod_distance_multiplier; @@ -586,7 +913,7 @@ void RenderForwardMobile::_render_shadow_append(RID p_framebuffer, const PagedAr _fill_render_list(RENDER_LIST_SECONDARY, &render_data, pass_mode, true); uint32_t render_list_size = render_list[RENDER_LIST_SECONDARY].elements.size() - render_list_from; render_list[RENDER_LIST_SECONDARY].sort_by_key_range(render_list_from, render_list_size); - _fill_instance_data(RENDER_LIST_SECONDARY, render_list_from, render_list_size, false); + _fill_element_info(RENDER_LIST_SECONDARY, render_list_from, render_list_size); { //regular forward for now @@ -631,7 +958,7 @@ void RenderForwardMobile::_render_shadow_end(uint32_t p_barrier) { for (uint32_t i = 0; i < scene_state.shadow_passes.size(); i++) { SceneState::ShadowPass &shadow_pass = scene_state.shadow_passes[i]; - RenderListParameters render_list_parameters(render_list[RENDER_LIST_SECONDARY].elements.ptr() + shadow_pass.element_from, render_list[RENDER_LIST_SECONDARY].element_info.ptr() + shadow_pass.element_from, shadow_pass.element_count, shadow_pass.flip_cull, shadow_pass.pass_mode, shadow_pass.rp_uniform_set, false, Vector2(), shadow_pass.camera_plane, shadow_pass.lod_distance_multiplier, shadow_pass.screen_lod_threshold, 1, shadow_pass.element_from, RD::BARRIER_MASK_NO_BARRIER); + RenderListParameters render_list_parameters(render_list[RENDER_LIST_SECONDARY].elements.ptr() + shadow_pass.element_from, render_list[RENDER_LIST_SECONDARY].element_info.ptr() + shadow_pass.element_from, shadow_pass.element_count, shadow_pass.flip_cull, shadow_pass.pass_mode, shadow_pass.rp_uniform_set, 0, false, Vector2(), shadow_pass.camera_plane, shadow_pass.lod_distance_multiplier, shadow_pass.screen_lod_threshold, 1, shadow_pass.element_from, RD::BARRIER_MASK_NO_BARRIER); _render_list_with_threads(&render_list_parameters, shadow_pass.framebuffer, RD::INITIAL_ACTION_DROP, RD::FINAL_ACTION_DISCARD, shadow_pass.initial_depth_action, shadow_pass.final_depth_action, Vector<Color>(), 1.0, 0, shadow_pass.rect); } @@ -664,14 +991,14 @@ void RenderForwardMobile::_render_material(const Transform3D &p_cam_transform, c PassMode pass_mode = PASS_MODE_DEPTH_MATERIAL; _fill_render_list(RENDER_LIST_SECONDARY, &render_data, pass_mode); render_list[RENDER_LIST_SECONDARY].sort_by_key(); - _fill_instance_data(RENDER_LIST_SECONDARY); + _fill_element_info(RENDER_LIST_SECONDARY); RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, nullptr, RID()); RENDER_TIMESTAMP("Render Material"); { - RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), true, pass_mode, rp_uniform_set); + RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), true, pass_mode, rp_uniform_set, 0); //regular forward for now Vector<Color> clear; clear.push_back(Color(0, 0, 0, 0)); @@ -705,14 +1032,14 @@ void RenderForwardMobile::_render_uv2(const PagedArray<GeometryInstance *> &p_in PassMode pass_mode = PASS_MODE_DEPTH_MATERIAL; _fill_render_list(RENDER_LIST_SECONDARY, &render_data, pass_mode); render_list[RENDER_LIST_SECONDARY].sort_by_key(); - _fill_instance_data(RENDER_LIST_SECONDARY); + _fill_element_info(RENDER_LIST_SECONDARY); RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, nullptr, RID()); RENDER_TIMESTAMP("Render Material"); { - RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), true, pass_mode, rp_uniform_set, true); + RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), true, pass_mode, rp_uniform_set, true, 0); //regular forward for now Vector<Color> clear; clear.push_back(Color(0, 0, 0, 0)); @@ -778,7 +1105,7 @@ void RenderForwardMobile::_render_particle_collider_heightfield(RID p_fb, const _fill_render_list(RENDER_LIST_SECONDARY, &render_data, pass_mode); render_list[RENDER_LIST_SECONDARY].sort_by_key(); - _fill_instance_data(RENDER_LIST_SECONDARY); + _fill_element_info(RENDER_LIST_SECONDARY); RID rp_uniform_set = _setup_render_pass_uniform_set(RENDER_LIST_SECONDARY, nullptr, RID()); @@ -786,7 +1113,7 @@ void RenderForwardMobile::_render_particle_collider_heightfield(RID p_fb, const { //regular forward for now - RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), false, pass_mode, rp_uniform_set); + RenderListParameters render_list_params(render_list[RENDER_LIST_SECONDARY].elements.ptr(), render_list[RENDER_LIST_SECONDARY].element_info.ptr(), render_list[RENDER_LIST_SECONDARY].elements.size(), false, pass_mode, rp_uniform_set, 0); _render_list_with_threads(&render_list_params, p_fb, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ); } RD::get_singleton()->draw_command_end_label(); @@ -843,13 +1170,67 @@ void RenderForwardMobile::_update_render_base_uniform_set() { { RD::Uniform u; u.binding = 3; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + RID sampler; + switch (decals_get_filter()) { + case RS::DECAL_FILTER_NEAREST: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::DECAL_FILTER_NEAREST_MIPMAPS: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::DECAL_FILTER_LINEAR: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::DECAL_FILTER_LINEAR_MIPMAPS: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::DECAL_FILTER_LINEAR_MIPMAPS_ANISOTROPIC: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + } + + u.ids.push_back(sampler); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 4; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER; + RID sampler; + switch (light_projectors_get_filter()) { + case RS::LIGHT_PROJECTOR_FILTER_NEAREST: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::LIGHT_PROJECTOR_FILTER_NEAREST_MIPMAPS: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::LIGHT_PROJECTOR_FILTER_LINEAR: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::LIGHT_PROJECTOR_FILTER_LINEAR_MIPMAPS: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + case RS::LIGHT_PROJECTOR_FILTER_LINEAR_MIPMAPS_ANISOTROPIC: { + sampler = storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED); + } break; + } + + u.ids.push_back(sampler); + uniforms.push_back(u); + } + + { + RD::Uniform u; + u.binding = 5; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.ids.push_back(get_omni_light_buffer()); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 4; + u.binding = 6; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.ids.push_back(get_spot_light_buffer()); uniforms.push_back(u); @@ -857,35 +1238,35 @@ void RenderForwardMobile::_update_render_base_uniform_set() { { RD::Uniform u; - u.binding = 5; + u.binding = 7; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.ids.push_back(get_reflection_probe_buffer()); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 6; + u.binding = 8; u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; u.ids.push_back(get_directional_light_buffer()); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 7; + u.binding = 9; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.ids.push_back(scene_state.lightmap_buffer); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 8; + u.binding = 10; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.ids.push_back(scene_state.lightmap_capture_buffer); uniforms.push_back(u); } { RD::Uniform u; - u.binding = 9; + u.binding = 11; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID decal_atlas = storage->decal_atlas_get_texture(); u.ids.push_back(decal_atlas); @@ -893,7 +1274,7 @@ void RenderForwardMobile::_update_render_base_uniform_set() { } { RD::Uniform u; - u.binding = 10; + u.binding = 12; u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; RID decal_atlas = storage->decal_atlas_get_texture_srgb(); u.ids.push_back(decal_atlas); @@ -901,7 +1282,7 @@ void RenderForwardMobile::_update_render_base_uniform_set() { } { RD::Uniform u; - u.binding = 11; + u.binding = 13; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; u.ids.push_back(get_decal_buffer()); uniforms.push_back(u); @@ -910,7 +1291,7 @@ void RenderForwardMobile::_update_render_base_uniform_set() { { RD::Uniform u; u.uniform_type = RD::UNIFORM_TYPE_STORAGE_BUFFER; - u.binding = 12; + u.binding = 14; u.ids.push_back(storage->global_variables_get_storage_buffer()); uniforms.push_back(u); } @@ -927,6 +1308,12 @@ RID RenderForwardMobile::_render_buffers_get_normal_texture(RID p_render_buffers return RID(); } +_FORCE_INLINE_ static uint32_t _indices_to_primitives(RS::PrimitiveType p_primitive, uint32_t p_indices) { + static const uint32_t divisor[RS::PRIMITIVE_MAX] = { 1, 2, 1, 3, 1 }; + static const uint32_t subtractor[RS::PRIMITIVE_MAX] = { 0, 0, 1, 0, 1 }; + return (p_indices - subtractor[p_primitive]) / divisor[p_primitive]; +} + void RenderForwardMobile::_fill_render_list(RenderListType p_render_list, const RenderDataRD *p_render_data, PassMode p_pass_mode, bool p_append) { if (p_render_list == RENDER_LIST_OPAQUE) { scene_state.used_sss = false; @@ -963,6 +1350,10 @@ void RenderForwardMobile::_fill_render_list(RenderListType p_render_list, const uint32_t flags = inst->base_flags; //fill flags if appropriate + if (inst->non_uniform_scale) { + flags |= INSTANCE_DATA_FLAGS_NON_UNIFORM_SCALE; + } + bool uses_lightmap = false; // bool uses_gi = false; @@ -1032,21 +1423,42 @@ void RenderForwardMobile::_fill_render_list(RenderListType p_render_list, const distance = -distance_max; } - surf->lod_index = storage->mesh_surface_get_lod(surf->surface, inst->lod_model_scale * inst->lod_bias, distance * p_render_data->lod_distance_multiplier, p_render_data->screen_lod_threshold); + uint32_t indices; + surf->lod_index = storage->mesh_surface_get_lod(surf->surface, inst->lod_model_scale * inst->lod_bias, distance * p_render_data->lod_distance_multiplier, p_render_data->screen_lod_threshold, &indices); + if (p_render_data->render_info) { + indices = _indices_to_primitives(surf->primitive, indices); + if (p_render_list == RENDER_LIST_OPAQUE) { //opaque + p_render_data->render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_VISIBLE][RS::VIEWPORT_RENDER_INFO_PRIMITIVES_IN_FRAME] += indices; + } else if (p_render_list == RENDER_LIST_SECONDARY) { //shadow + p_render_data->render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_SHADOW][RS::VIEWPORT_RENDER_INFO_PRIMITIVES_IN_FRAME] += indices; + } + } } else { surf->lod_index = 0; + if (p_render_data->render_info) { + uint32_t to_draw = storage->mesh_surface_get_vertices_drawn_count(surf->surface); + to_draw = _indices_to_primitives(surf->primitive, to_draw); + to_draw *= inst->instance_count; + if (p_render_list == RENDER_LIST_OPAQUE) { //opaque + p_render_data->render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_VISIBLE][RS::VIEWPORT_RENDER_INFO_PRIMITIVES_IN_FRAME] += storage->mesh_surface_get_vertices_drawn_count(surf->surface); + } else if (p_render_list == RENDER_LIST_SECONDARY) { //shadow + p_render_data->render_info->info[RS::VIEWPORT_RENDER_INFO_TYPE_SHADOW][RS::VIEWPORT_RENDER_INFO_PRIMITIVES_IN_FRAME] += storage->mesh_surface_get_vertices_drawn_count(surf->surface); + } + } } // ADD Element if (p_pass_mode == PASS_MODE_COLOR) { - if (surf->flags & (GeometryInstanceSurfaceDataCache::FLAG_PASS_DEPTH | GeometryInstanceSurfaceDataCache::FLAG_PASS_OPAQUE)) { +#ifdef DEBUG_ENABLED + bool force_alpha = unlikely(get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_OVERDRAW); +#else + bool force_alpha = false; +#endif + if (!force_alpha && (surf->flags & (GeometryInstanceSurfaceDataCache::FLAG_PASS_DEPTH | GeometryInstanceSurfaceDataCache::FLAG_PASS_OPAQUE))) { rl->add_element(surf); } - if (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_PASS_ALPHA) { + if (force_alpha || (surf->flags & GeometryInstanceSurfaceDataCache::FLAG_PASS_ALPHA)) { render_list[RENDER_LIST_ALPHA].add_element(surf); - // if (uses_gi) { - // surf->sort.uses_forward_gi = 1; - // } } if (uses_lightmap) { @@ -1116,11 +1528,6 @@ void RenderForwardMobile::_setup_environment(const RenderDataRD *p_render_data, RendererStorageRD::store_soft_shadow_kernel(penumbra_shadow_kernel_get(), scene_state.ubo.penumbra_shadow_kernel); RendererStorageRD::store_soft_shadow_kernel(soft_shadow_kernel_get(), scene_state.ubo.soft_shadow_kernel); - scene_state.ubo.directional_penumbra_shadow_samples = directional_penumbra_shadow_samples_get(); - scene_state.ubo.directional_soft_shadow_samples = directional_soft_shadow_samples_get(); - scene_state.ubo.penumbra_shadow_samples = penumbra_shadow_samples_get(); - scene_state.ubo.soft_shadow_samples = soft_shadow_samples_get(); - Size2 screen_pixel_size = Vector2(1.0, 1.0) / Size2(p_screen_size); scene_state.ubo.screen_pixel_size[0] = screen_pixel_size.x; scene_state.ubo.screen_pixel_size[1] = screen_pixel_size.y; @@ -1284,9 +1691,7 @@ void RenderForwardMobile::_setup_environment(const RenderDataRD *p_render_data, RD::get_singleton()->buffer_update(scene_state.uniform_buffers[p_index], 0, sizeof(SceneState::UBO), &scene_state.ubo, RD::BARRIER_MASK_RASTER); } -void RenderForwardMobile::_fill_instance_data(RenderListType p_render_list, uint32_t p_offset, int32_t p_max_elements, bool p_update_buffer) { - // !BAS! Rename this to make clear this is not the same as with the forward renderer and remove p_update_buffer? - +void RenderForwardMobile::_fill_element_info(RenderListType p_render_list, uint32_t p_offset, int32_t p_max_elements) { RenderList *rl = &render_list[p_render_list]; uint32_t element_total = p_max_elements >= 0 ? uint32_t(p_max_elements) : rl->elements.size(); @@ -1351,6 +1756,57 @@ void RenderForwardMobile::_render_list_with_threads(RenderListParameters *p_para } } +void RenderForwardMobile::_fill_push_constant_instance_indices(GeometryInstanceForwardMobile::PushConstant *p_push_constant, uint32_t &spec_constants, const GeometryInstanceForwardMobile *p_instance) { + // first zero out our indices + + p_push_constant->omni_lights[0] = 0xFFFF; + p_push_constant->omni_lights[1] = 0xFFFF; + + p_push_constant->spot_lights[0] = 0xFFFF; + p_push_constant->spot_lights[1] = 0xFFFF; + + p_push_constant->decals[0] = 0xFFFF; + p_push_constant->decals[1] = 0xFFFF; + + p_push_constant->reflection_probes[0] = 0xFFFF; + p_push_constant->reflection_probes[1] = 0xFFFF; + + if (p_instance->omni_light_count == 0) { + spec_constants |= 1 << SPEC_CONSTANT_DISABLE_OMNI_LIGHTS; + } + if (p_instance->spot_light_count == 0) { + spec_constants |= 1 << SPEC_CONSTANT_DISABLE_SPOT_LIGHTS; + } + if (p_instance->reflection_probe_count == 0) { + spec_constants |= 1 << SPEC_CONSTANT_DISABLE_REFLECTION_PROBES; + } + if (p_instance->decals_count == 0) { + spec_constants |= 1 << SPEC_CONSTANT_DISABLE_DECALS; + } + + for (uint32_t i = 0; i < MAX_RDL_CULL; i++) { + uint32_t ofs = i < 4 ? 0 : 1; + uint32_t shift = (i & 0x3) << 3; + uint32_t mask = ~(0xFF << shift); + if (i < p_instance->omni_light_count) { + p_push_constant->omni_lights[ofs] &= mask; + p_push_constant->omni_lights[ofs] |= uint32_t(forward_id_allocators[FORWARD_ID_TYPE_OMNI_LIGHT].map[p_instance->omni_lights[i]]) << shift; + } + if (i < p_instance->spot_light_count) { + p_push_constant->spot_lights[ofs] &= mask; + p_push_constant->spot_lights[ofs] |= uint32_t(forward_id_allocators[FORWARD_ID_TYPE_SPOT_LIGHT].map[p_instance->spot_lights[i]]) << shift; + } + if (i < p_instance->decals_count) { + p_push_constant->decals[ofs] &= mask; + p_push_constant->decals[ofs] |= uint32_t(forward_id_allocators[FORWARD_ID_TYPE_DECAL].map[p_instance->decals[i]]) << shift; + } + if (i < p_instance->reflection_probe_count) { + p_push_constant->reflection_probes[ofs] &= mask; + p_push_constant->reflection_probes[ofs] |= uint32_t(forward_id_allocators[FORWARD_ID_TYPE_REFLECTION_PROBE].map[p_instance->reflection_probes[i]]) << shift; + } + } +} + template <RenderForwardMobile::PassMode p_pass_mode> void RenderForwardMobile::_render_list_template(RenderingDevice::DrawListID p_draw_list, RenderingDevice::FramebufferFormatID p_framebuffer_Format, RenderListParameters *p_params, uint32_t p_from_element, uint32_t p_to_element) { RD::DrawListID draw_list = p_draw_list; @@ -1375,6 +1831,8 @@ void RenderForwardMobile::_render_list_template(RenderingDevice::DrawListID p_dr const RenderElementInfo &element_info = p_params->element_info[i]; const GeometryInstanceForwardMobile *inst = surf->owner; + uint32_t base_spec_constants = p_params->spec_constant_base_flags; + // GeometryInstanceForwardMobile::PushConstant push_constant = inst->push_constant; GeometryInstanceForwardMobile::PushConstant push_constant; @@ -1400,8 +1858,6 @@ void RenderForwardMobile::_render_list_template(RenderingDevice::DrawListID p_dr push_constant.lightmap_uv_scale[3] = inst->lightmap_uv_scale.size.y; }; - _fill_instance_indices(inst->omni_lights, inst->omni_light_count, push_constant.omni_lights, inst->spot_lights, inst->spot_light_count, push_constant.spot_lights, inst->reflection_probes, inst->reflection_probe_count, push_constant.reflection_probes, inst->decals, inst->decals_count, push_constant.decals, push_constant.layer_mask); - RID material_uniform_set; SceneShaderForwardMobile::ShaderData *shader; void *mesh_surface; @@ -1412,8 +1868,28 @@ void RenderForwardMobile::_render_list_template(RenderingDevice::DrawListID p_dr mesh_surface = surf->surface_shadow; } else { - material_uniform_set = surf->material_uniform_set; - shader = surf->shader; + if (inst->use_projector) { + base_spec_constants |= 1 << SPEC_CONSTANT_USING_PROJECTOR; + } + if (inst->use_soft_shadow) { + base_spec_constants |= 1 << SPEC_CONSTANT_USING_SOFT_SHADOWS; + } + _fill_push_constant_instance_indices(&push_constant, base_spec_constants, inst); + +#ifdef DEBUG_ENABLED + if (unlikely(get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_LIGHTING)) { + material_uniform_set = scene_shader.default_material_uniform_set; + shader = scene_shader.default_material_shader_ptr; + } else if (unlikely(get_debug_draw_mode() == RS::VIEWPORT_DEBUG_DRAW_OVERDRAW)) { + material_uniform_set = scene_shader.overdraw_material_uniform_set; + shader = scene_shader.overdraw_material_shader_ptr; + } else { +#endif + material_uniform_set = surf->material_uniform_set; + shader = surf->shader; +#ifdef DEBUG_ENABLED + } +#endif mesh_surface = surf->surface; } @@ -1490,7 +1966,7 @@ void RenderForwardMobile::_render_list_template(RenderingDevice::DrawListID p_dr prev_index_array_rd = index_array_rd; } - RID pipeline_rd = pipeline->get_render_pipeline(vertex_format, framebuffer_format, p_params->force_wireframe); + RID pipeline_rd = pipeline->get_render_pipeline(vertex_format, framebuffer_format, p_params->force_wireframe, p_params->subpass, base_spec_constants); if (pipeline_rd != prev_pipeline_rd) { // checking with prev shader does not make so much sense, as @@ -1505,8 +1981,8 @@ void RenderForwardMobile::_render_list_template(RenderingDevice::DrawListID p_dr } if (material_uniform_set != prev_material_uniform_set) { - //update uniform set - if (material_uniform_set.is_valid()) { + // Update uniform set. + if (material_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(material_uniform_set)) { // Material may not have a uniform set. RD::get_singleton()->draw_list_bind_uniform_set(draw_list, material_uniform_set, MATERIAL_UNIFORM_SET); } @@ -1708,13 +2184,13 @@ void RenderForwardMobile::geometry_instance_pair_light_instances(GeometryInstanc switch (type) { case RS::LIGHT_OMNI: { if (ginstance->omni_light_count < (uint32_t)MAX_RDL_CULL) { - ginstance->omni_lights[ginstance->omni_light_count] = p_light_instances[i]; + ginstance->omni_lights[ginstance->omni_light_count] = light_instance_get_forward_id(p_light_instances[i]); ginstance->omni_light_count++; } } break; case RS::LIGHT_SPOT: { if (ginstance->spot_light_count < (uint32_t)MAX_RDL_CULL) { - ginstance->spot_lights[ginstance->spot_light_count] = p_light_instances[i]; + ginstance->spot_lights[ginstance->spot_light_count] = light_instance_get_forward_id(p_light_instances[i]); ginstance->spot_light_count++; } } break; @@ -1730,7 +2206,7 @@ void RenderForwardMobile::geometry_instance_pair_reflection_probe_instances(Geom ginstance->reflection_probe_count = p_reflection_probe_instance_count < (uint32_t)MAX_RDL_CULL ? p_reflection_probe_instance_count : (uint32_t)MAX_RDL_CULL; for (uint32_t i = 0; i < ginstance->reflection_probe_count; i++) { - ginstance->reflection_probes[i] = p_reflection_probe_instances[i]; + ginstance->reflection_probes[i] = reflection_probe_instance_get_forward_id(p_reflection_probe_instances[i]); } } @@ -1740,7 +2216,7 @@ void RenderForwardMobile::geometry_instance_pair_decal_instances(GeometryInstanc ginstance->decals_count = p_decal_instance_count < (uint32_t)MAX_RDL_CULL ? p_decal_instance_count : (uint32_t)MAX_RDL_CULL; for (uint32_t i = 0; i < ginstance->decals_count; i++) { - ginstance->decals[i] = p_decal_instances[i]; + ginstance->decals[i] = decal_instance_get_forward_id(p_decal_instances[i]); } } @@ -1748,6 +2224,14 @@ void RenderForwardMobile::geometry_instance_pair_voxel_gi_instances(GeometryInst // We do not have this here! } +void RenderForwardMobile::geometry_instance_set_softshadow_projector_pairing(GeometryInstance *p_geometry_instance, bool p_softshadow, bool p_projector) { + GeometryInstanceForwardMobile *ginstance = static_cast<GeometryInstanceForwardMobile *>(p_geometry_instance); + ERR_FAIL_COND(!ginstance); + + ginstance->use_projector = p_projector; + ginstance->use_soft_shadow = p_softshadow; +} + void RenderForwardMobile::_geometry_instance_mark_dirty(GeometryInstance *p_geometry_instance) { GeometryInstanceForwardMobile *ginstance = static_cast<GeometryInstanceForwardMobile *>(p_geometry_instance); if (ginstance->dirty_list_element.in_list()) { @@ -1995,6 +2479,7 @@ void RenderForwardMobile::_geometry_instance_update(GeometryInstance *p_geometry //Fill push constant bool store_transform = true; + ginstance->base_flags = 0; if (ginstance->data->base_type == RS::INSTANCE_MULTIMESH) { ginstance->base_flags |= INSTANCE_DATA_FLAG_MULTIMESH; @@ -2096,10 +2581,54 @@ uint32_t RenderForwardMobile::get_max_elements() const { RenderForwardMobile *RenderForwardMobile::singleton = nullptr; +void RenderForwardMobile::_update_shader_quality_settings() { + Vector<RD::PipelineSpecializationConstant> spec_constants; + + RD::PipelineSpecializationConstant sc; + sc.type = RD::PIPELINE_SPECIALIZATION_CONSTANT_TYPE_INT; + + sc.constant_id = SPEC_CONSTANT_SOFT_SHADOW_SAMPLES; + sc.int_value = soft_shadow_samples_get(); + + spec_constants.push_back(sc); + + sc.constant_id = SPEC_CONSTANT_PENUMBRA_SHADOW_SAMPLES; + sc.int_value = penumbra_shadow_samples_get(); + + spec_constants.push_back(sc); + + sc.constant_id = SPEC_CONSTANT_DIRECTIONAL_SOFT_SHADOW_SAMPLES; + sc.int_value = directional_soft_shadow_samples_get(); + + spec_constants.push_back(sc); + + sc.constant_id = SPEC_CONSTANT_DIRECTIONAL_PENUMBRA_SHADOW_SAMPLES; + sc.int_value = directional_penumbra_shadow_samples_get(); + + spec_constants.push_back(sc); + + sc.type = RD::PIPELINE_SPECIALIZATION_CONSTANT_TYPE_BOOL; + sc.constant_id = SPEC_CONSTANT_DECAL_USE_MIPMAPS; + sc.bool_value = decals_get_filter() == RS::DECAL_FILTER_NEAREST_MIPMAPS || decals_get_filter() == RS::DECAL_FILTER_LINEAR_MIPMAPS || decals_get_filter() == RS::DECAL_FILTER_LINEAR_MIPMAPS_ANISOTROPIC; + + spec_constants.push_back(sc); + + sc.constant_id = SPEC_CONSTANT_PROJECTOR_USE_MIPMAPS; + sc.bool_value = light_projectors_get_filter() == RS::LIGHT_PROJECTOR_FILTER_NEAREST_MIPMAPS || light_projectors_get_filter() == RS::LIGHT_PROJECTOR_FILTER_LINEAR_MIPMAPS || light_projectors_get_filter() == RS::LIGHT_PROJECTOR_FILTER_LINEAR_MIPMAPS_ANISOTROPIC; + + spec_constants.push_back(sc); + + scene_shader.set_default_specialization_constants(spec_constants); + + _base_uniforms_changed(); //also need this +} + RenderForwardMobile::RenderForwardMobile(RendererStorageRD *p_storage) : RendererSceneRenderRD(p_storage) { singleton = this; + sky.set_texture_format(_render_buffers_get_color_format()); + String defines; defines += "\n#define MAX_ROUGHNESS_LOD " + itos(get_roughness_layers() - 1) + ".0\n"; @@ -2107,7 +2636,7 @@ RenderForwardMobile::RenderForwardMobile(RendererStorageRD *p_storage) : defines += "\n#define USE_RADIANCE_CUBEMAP_ARRAY \n"; } // defines += "\n#define SDFGI_OCT_SIZE " + itos(gi.sdfgi_get_lightprobe_octahedron_size()) + "\n"; - defines += "\n#define MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS " + itos(get_max_directional_lights()) + "\n"; + defines += "\n#define MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS " + itos(MAX_DIRECTIONAL_LIGHTS) + "\n"; { //lightmaps @@ -2131,6 +2660,8 @@ RenderForwardMobile::RenderForwardMobile(RendererStorageRD *p_storage) : // !BAS! maybe we need a mobile version of this setting? render_list_thread_threshold = GLOBAL_GET("rendering/limits/forward_renderer/threaded_render_minimum_instances"); + + _update_shader_quality_settings(); } RenderForwardMobile::~RenderForwardMobile() { diff --git a/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.h b/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.h index 99cbd45b10..38f80c5347 100644 --- a/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.h +++ b/servers/rendering/renderer_rd/forward_mobile/render_forward_mobile.h @@ -42,6 +42,18 @@ namespace RendererSceneRenderImplementation { class RenderForwardMobile : public RendererSceneRenderRD { friend SceneShaderForwardMobile; + struct ForwardIDAllocator { + LocalVector<bool> allocations; + LocalVector<uint8_t> map; + }; + + ForwardIDAllocator forward_id_allocators[FORWARD_ID_MAX]; + + virtual ForwardID _allocate_forward_id(ForwardIDType p_type) override; + virtual void _free_forward_id(ForwardIDType p_type, ForwardID p_id) override; + virtual void _map_forward_id(ForwardIDType p_type, ForwardID p_id, uint32_t p_index) override; + virtual bool _uses_forward_ids() const override { return true; } + protected: /* Scene Shader */ @@ -53,6 +65,30 @@ protected: }; enum { + + SPEC_CONSTANT_USING_PROJECTOR = 0, + SPEC_CONSTANT_USING_SOFT_SHADOWS = 1, + SPEC_CONSTANT_USING_DIRECTIONAL_SOFT_SHADOWS = 2, + + SPEC_CONSTANT_SOFT_SHADOW_SAMPLES = 3, + SPEC_CONSTANT_PENUMBRA_SHADOW_SAMPLES = 4, + SPEC_CONSTANT_DIRECTIONAL_SOFT_SHADOW_SAMPLES = 5, + SPEC_CONSTANT_DIRECTIONAL_PENUMBRA_SHADOW_SAMPLES = 6, + + SPEC_CONSTANT_DECAL_USE_MIPMAPS = 7, + SPEC_CONSTANT_PROJECTOR_USE_MIPMAPS = 8, + + SPEC_CONSTANT_DISABLE_OMNI_LIGHTS = 9, + SPEC_CONSTANT_DISABLE_SPOT_LIGHTS = 10, + SPEC_CONSTANT_DISABLE_REFLECTION_PROBES = 11, + SPEC_CONSTANT_DISABLE_DIRECTIONAL_LIGHTS = 12, + + SPEC_CONSTANT_DISABLE_DECALS = 13, + SPEC_CONSTANT_DISABLE_FOG = 14, + + }; + + enum { MAX_LIGHTMAPS = 8, MAX_RDL_CULL = 8, // maximum number of reflection probes, decals or lights we can cull per geometry instance INSTANCE_DATA_BUFFER_MIN_SIZE = 4096 @@ -71,6 +107,18 @@ protected: /* Render Buffer */ + // We can have: + // - 4 subpasses combining the full render cycle + // - 3 subpasses + 1 normal pass for tonemapping/glow/dof/etc (using fb for 2D buffer) + // - 2 subpasses + 1 normal pass for transparent + 1 normal pass for tonemapping/glow/dof/etc (using fb for 2D buffer) + enum RenderBufferMobileFramebufferConfigType { + FB_CONFIG_ONE_PASS, // Single pass frame buffer for alpha pass + FB_CONFIG_TWO_SUBPASSES, // Opaque + Sky sub pass + FB_CONFIG_THREE_SUBPASSES, // Opaque + Sky + Alpha sub pass + FB_CONFIG_FOUR_SUBPASSES, // Opaque + Sky + Alpha sub pass + Tonemap pass + FB_CONFIG_MAX + }; + struct RenderBufferDataForwardMobile : public RenderBufferData { RID color; RID depth; @@ -83,17 +131,17 @@ protected: RID depth_msaa; // RID normal_roughness_buffer_msaa; - RID color_fb; + RID color_fbs[FB_CONFIG_MAX]; int width, height; uint32_t view_count; void clear(); - virtual void configure(RID p_color_buffer, RID p_depth_buffer, int p_width, int p_height, RS::ViewportMSAA p_msaa, uint32_t p_view_count); + virtual void configure(RID p_color_buffer, RID p_depth_buffer, RID p_target_buffer, int p_width, int p_height, RS::ViewportMSAA p_msaa, uint32_t p_view_count); ~RenderBufferDataForwardMobile(); }; - virtual RenderBufferData *_create_render_buffer_data(); + virtual RenderBufferData *_create_render_buffer_data() override; /* Rendering */ @@ -126,13 +174,15 @@ protected: bool force_wireframe = false; Vector2 uv_offset; Plane lod_plane; + uint32_t spec_constant_base_flags = 0; float lod_distance_multiplier = 0.0; float screen_lod_threshold = 0.0; RD::FramebufferFormatID framebuffer_format = 0; uint32_t element_offset = 0; uint32_t barrier = RD::BARRIER_MASK_ALL; + uint32_t subpass = 0; - RenderListParameters(GeometryInstanceSurfaceDataCache **p_elements, RenderElementInfo *p_element_info, int p_element_count, bool p_reverse_cull, PassMode p_pass_mode, RID p_render_pass_uniform_set, bool p_force_wireframe = false, const Vector2 &p_uv_offset = Vector2(), const Plane &p_lod_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0, uint32_t p_view_count = 1, uint32_t p_element_offset = 0, uint32_t p_barrier = RD::BARRIER_MASK_ALL) { + RenderListParameters(GeometryInstanceSurfaceDataCache **p_elements, RenderElementInfo *p_element_info, int p_element_count, bool p_reverse_cull, PassMode p_pass_mode, RID p_render_pass_uniform_set, uint32_t p_spec_constant_base_flags = 0, bool p_force_wireframe = false, const Vector2 &p_uv_offset = Vector2(), const Plane &p_lod_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0, uint32_t p_view_count = 1, uint32_t p_element_offset = 0, uint32_t p_barrier = RD::BARRIER_MASK_ALL) { elements = p_elements; element_info = p_element_info; element_count = p_element_count; @@ -148,30 +198,35 @@ protected: screen_lod_threshold = p_screen_lod_threshold; element_offset = p_element_offset; barrier = p_barrier; + spec_constant_base_flags = p_spec_constant_base_flags; } }; + virtual float _render_buffers_get_luminance_multiplier() override; + virtual RD::DataFormat _render_buffers_get_color_format() override; + virtual bool _render_buffers_can_be_storage() override; + RID _setup_render_pass_uniform_set(RenderListType p_render_list, const RenderDataRD *p_render_data, RID p_radiance_texture, bool p_use_directional_shadow_atlas = false, int p_index = 0); - virtual void _render_scene(RenderDataRD *p_render_data, const Color &p_default_bg_color); + virtual void _render_scene(RenderDataRD *p_render_data, const Color &p_default_bg_color) override; - virtual void _render_shadow_begin(); - virtual void _render_shadow_append(RID p_framebuffer, const PagedArray<GeometryInstance *> &p_instances, const CameraMatrix &p_projection, const Transform3D &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0, const Rect2i &p_rect = Rect2i(), bool p_flip_y = false, bool p_clear_region = true, bool p_begin = true, bool p_end = true); - virtual void _render_shadow_process(); - virtual void _render_shadow_end(uint32_t p_barrier = RD::BARRIER_MASK_ALL); + virtual void _render_shadow_begin() override; + virtual void _render_shadow_append(RID p_framebuffer, const PagedArray<GeometryInstance *> &p_instances, const CameraMatrix &p_projection, const Transform3D &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0, const Rect2i &p_rect = Rect2i(), bool p_flip_y = false, bool p_clear_region = true, bool p_begin = true, bool p_end = true, RendererScene::RenderInfo *p_render_info = nullptr) override; + virtual void _render_shadow_process() override; + virtual void _render_shadow_end(uint32_t p_barrier = RD::BARRIER_MASK_ALL) override; - virtual void _render_material(const Transform3D &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region); - virtual void _render_uv2(const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region); - virtual void _render_sdfgi(RID p_render_buffers, const Vector3i &p_from, const Vector3i &p_size, const AABB &p_bounds, const PagedArray<GeometryInstance *> &p_instances, const RID &p_albedo_texture, const RID &p_emission_texture, const RID &p_emission_aniso_texture, const RID &p_geom_facing_texture); - virtual void _render_particle_collider_heightfield(RID p_fb, const Transform3D &p_cam_transform, const CameraMatrix &p_cam_projection, const PagedArray<GeometryInstance *> &p_instances); + virtual void _render_material(const Transform3D &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) override; + virtual void _render_uv2(const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) override; + virtual void _render_sdfgi(RID p_render_buffers, const Vector3i &p_from, const Vector3i &p_size, const AABB &p_bounds, const PagedArray<GeometryInstance *> &p_instances, const RID &p_albedo_texture, const RID &p_emission_texture, const RID &p_emission_aniso_texture, const RID &p_geom_facing_texture) override; + virtual void _render_particle_collider_heightfield(RID p_fb, const Transform3D &p_cam_transform, const CameraMatrix &p_cam_projection, const PagedArray<GeometryInstance *> &p_instances) override; uint64_t lightmap_texture_array_version = 0xFFFFFFFF; - virtual void _base_uniforms_changed(); + virtual void _base_uniforms_changed() override; void _update_render_base_uniform_set(); - virtual RID _render_buffers_get_normal_texture(RID p_render_buffers); + virtual RID _render_buffers_get_normal_texture(RID p_render_buffers) override; void _fill_render_list(RenderListType p_render_list, const RenderDataRD *p_render_data, PassMode p_pass_mode, bool p_append = false); - void _fill_instance_data(RenderListType p_render_list, uint32_t p_offset = 0, int32_t p_max_elements = -1, bool p_update_buffer = true); + void _fill_element_info(RenderListType p_render_list, uint32_t p_offset = 0, int32_t p_max_elements = -1); // void _update_instance_data_buffer(RenderListType p_render_list); static RenderForwardMobile *singleton; @@ -213,11 +268,6 @@ protected: float penumbra_shadow_kernel[128]; float soft_shadow_kernel[128]; - uint32_t directional_penumbra_shadow_samples; - uint32_t directional_soft_shadow_samples; - uint32_t penumbra_shadow_samples; - uint32_t soft_shadow_samples; - float ambient_light_color_energy[4]; float ambient_color_sky_mix; @@ -356,7 +406,7 @@ protected: } }; - void sort_by_reverse_depth_and_priority(bool p_alpha) { //used for alpha + void sort_by_reverse_depth_and_priority() { //used for alpha SortArray<GeometryInstanceSurfaceDataCache *, SortByReverseDepthAndPriority> sorter; sorter.sort(elements.ptr(), elements.size()); @@ -390,6 +440,7 @@ protected: // check which ones of these apply, probably all except GI and SDFGI enum { + INSTANCE_DATA_FLAGS_NON_UNIFORM_SCALE = 1 << 5, INSTANCE_DATA_FLAG_USE_GI_BUFFERS = 1 << 6, INSTANCE_DATA_FLAG_USE_SDFGI = 1 << 7, INSTANCE_DATA_FLAG_USE_LIGHTMAP_CAPTURE = 1 << 8, @@ -402,7 +453,6 @@ protected: INSTANCE_DATA_FLAG_MULTIMESH_HAS_CUSTOM_DATA = 1 << 15, INSTANCE_DATA_FLAGS_PARTICLE_TRAIL_SHIFT = 16, INSTANCE_DATA_FLAGS_PARTICLE_TRAIL_MASK = 0xFF, - INSTANCE_DATA_FLAGS_NON_UNIFORM_SCALE = 1 << 24, }; struct GeometryInstanceLightmapSH { @@ -494,6 +544,8 @@ protected: RID transforms_uniform_set; float depth = 0; bool mirror = false; + bool use_projector = false; + bool use_soft_shadow = false; Transform3D transform; bool store_transform_cache = true; // if true we copy our transform into our PushConstant, if false we use our transforms UBO and clear our PushConstants transform bool non_uniform_scale = false; @@ -513,14 +565,14 @@ protected: GeometryInstanceLightmapSH *lightmap_sh = nullptr; // culled light info - uint32_t reflection_probe_count; - RID reflection_probes[MAX_RDL_CULL]; - uint32_t omni_light_count; - RID omni_lights[MAX_RDL_CULL]; - uint32_t spot_light_count; - RID spot_lights[MAX_RDL_CULL]; - uint32_t decals_count; - RID decals[MAX_RDL_CULL]; + uint32_t reflection_probe_count = 0; + ForwardID reflection_probes[MAX_RDL_CULL]; + uint32_t omni_light_count = 0; + ForwardID omni_lights[MAX_RDL_CULL]; + uint32_t spot_light_count = 0; + ForwardID spot_lights[MAX_RDL_CULL]; + uint32_t decals_count = 0; + ForwardID decals[MAX_RDL_CULL]; GeometryInstanceSurfaceDataCache *surface_caches = nullptr; @@ -552,7 +604,13 @@ protected: dirty_list_element(this) {} }; + _FORCE_INLINE_ void _fill_push_constant_instance_indices(GeometryInstanceForwardMobile::PushConstant *p_push_constant, uint32_t &spec_constants, const GeometryInstanceForwardMobile *p_instance); + + void _update_shader_quality_settings() override; + public: + virtual RID reflection_probe_create_framebuffer(RID p_color, RID p_depth) override; + static void _geometry_instance_dependency_changed(RendererStorage::DependencyChangedNotification p_notification, RendererStorage::DependencyTracker *p_tracker); static void _geometry_instance_dependency_deleted(const RID &p_dependency, RendererStorage::DependencyTracker *p_tracker); @@ -568,38 +626,40 @@ public: void _geometry_instance_update(GeometryInstance *p_geometry_instance); void _update_dirty_geometry_instances(); - virtual GeometryInstance *geometry_instance_create(RID p_base); - virtual void geometry_instance_set_skeleton(GeometryInstance *p_geometry_instance, RID p_skeleton); - virtual void geometry_instance_set_material_override(GeometryInstance *p_geometry_instance, RID p_override); - virtual void geometry_instance_set_surface_materials(GeometryInstance *p_geometry_instance, const Vector<RID> &p_materials); - virtual void geometry_instance_set_mesh_instance(GeometryInstance *p_geometry_instance, RID p_mesh_instance); - virtual void geometry_instance_set_transform(GeometryInstance *p_geometry_instance, const Transform3D &p_transform, const AABB &p_aabb, const AABB &p_transformed_aabb); - virtual void geometry_instance_set_layer_mask(GeometryInstance *p_geometry_instance, uint32_t p_layer_mask); - virtual void geometry_instance_set_lod_bias(GeometryInstance *p_geometry_instance, float p_lod_bias); - virtual void geometry_instance_set_use_baked_light(GeometryInstance *p_geometry_instance, bool p_enable); - virtual void geometry_instance_set_use_dynamic_gi(GeometryInstance *p_geometry_instance, bool p_enable); - virtual void geometry_instance_set_use_lightmap(GeometryInstance *p_geometry_instance, RID p_lightmap_instance, const Rect2 &p_lightmap_uv_scale, int p_lightmap_slice_index); - virtual void geometry_instance_set_lightmap_capture(GeometryInstance *p_geometry_instance, const Color *p_sh9); - virtual void geometry_instance_set_instance_shader_parameters_offset(GeometryInstance *p_geometry_instance, int32_t p_offset); - virtual void geometry_instance_set_cast_double_sided_shadows(GeometryInstance *p_geometry_instance, bool p_enable); - - virtual Transform3D geometry_instance_get_transform(GeometryInstance *p_instance); - virtual AABB geometry_instance_get_aabb(GeometryInstance *p_instance); - - virtual void geometry_instance_free(GeometryInstance *p_geometry_instance); - - virtual uint32_t geometry_instance_get_pair_mask(); - virtual void geometry_instance_pair_light_instances(GeometryInstance *p_geometry_instance, const RID *p_light_instances, uint32_t p_light_instance_count); - virtual void geometry_instance_pair_reflection_probe_instances(GeometryInstance *p_geometry_instance, const RID *p_reflection_probe_instances, uint32_t p_reflection_probe_instance_count); - virtual void geometry_instance_pair_decal_instances(GeometryInstance *p_geometry_instance, const RID *p_decal_instances, uint32_t p_decal_instance_count); - virtual void geometry_instance_pair_voxel_gi_instances(GeometryInstance *p_geometry_instance, const RID *p_voxel_gi_instances, uint32_t p_voxel_gi_instance_count); - - virtual bool free(RID p_rid); - - virtual bool is_dynamic_gi_supported() const; - virtual bool is_clustered_enabled() const; - virtual bool is_volumetric_supported() const; - virtual uint32_t get_max_elements() const; + virtual GeometryInstance *geometry_instance_create(RID p_base) override; + virtual void geometry_instance_set_skeleton(GeometryInstance *p_geometry_instance, RID p_skeleton) override; + virtual void geometry_instance_set_material_override(GeometryInstance *p_geometry_instance, RID p_override) override; + virtual void geometry_instance_set_surface_materials(GeometryInstance *p_geometry_instance, const Vector<RID> &p_materials) override; + virtual void geometry_instance_set_mesh_instance(GeometryInstance *p_geometry_instance, RID p_mesh_instance) override; + virtual void geometry_instance_set_transform(GeometryInstance *p_geometry_instance, const Transform3D &p_transform, const AABB &p_aabb, const AABB &p_transformed_aabb) override; + virtual void geometry_instance_set_layer_mask(GeometryInstance *p_geometry_instance, uint32_t p_layer_mask) override; + virtual void geometry_instance_set_lod_bias(GeometryInstance *p_geometry_instance, float p_lod_bias) override; + virtual void geometry_instance_set_use_baked_light(GeometryInstance *p_geometry_instance, bool p_enable) override; + virtual void geometry_instance_set_use_dynamic_gi(GeometryInstance *p_geometry_instance, bool p_enable) override; + virtual void geometry_instance_set_use_lightmap(GeometryInstance *p_geometry_instance, RID p_lightmap_instance, const Rect2 &p_lightmap_uv_scale, int p_lightmap_slice_index) override; + virtual void geometry_instance_set_lightmap_capture(GeometryInstance *p_geometry_instance, const Color *p_sh9) override; + virtual void geometry_instance_set_instance_shader_parameters_offset(GeometryInstance *p_geometry_instance, int32_t p_offset) override; + virtual void geometry_instance_set_cast_double_sided_shadows(GeometryInstance *p_geometry_instance, bool p_enable) override; + + virtual Transform3D geometry_instance_get_transform(GeometryInstance *p_instance) override; + virtual AABB geometry_instance_get_aabb(GeometryInstance *p_instance) override; + + virtual void geometry_instance_free(GeometryInstance *p_geometry_instance) override; + + virtual uint32_t geometry_instance_get_pair_mask() override; + virtual void geometry_instance_pair_light_instances(GeometryInstance *p_geometry_instance, const RID *p_light_instances, uint32_t p_light_instance_count) override; + virtual void geometry_instance_pair_reflection_probe_instances(GeometryInstance *p_geometry_instance, const RID *p_reflection_probe_instances, uint32_t p_reflection_probe_instance_count) override; + virtual void geometry_instance_pair_decal_instances(GeometryInstance *p_geometry_instance, const RID *p_decal_instances, uint32_t p_decal_instance_count) override; + virtual void geometry_instance_pair_voxel_gi_instances(GeometryInstance *p_geometry_instance, const RID *p_voxel_gi_instances, uint32_t p_voxel_gi_instance_count) override; + + virtual void geometry_instance_set_softshadow_projector_pairing(GeometryInstance *p_geometry_instance, bool p_softshadow, bool p_projector) override; + + virtual bool free(RID p_rid) override; + + virtual bool is_dynamic_gi_supported() const override; + virtual bool is_clustered_enabled() const override; + virtual bool is_volumetric_supported() const override; + virtual uint32_t get_max_elements() const override; RenderForwardMobile(RendererStorageRD *p_storage); ~RenderForwardMobile(); diff --git a/servers/rendering/renderer_rd/forward_mobile/scene_shader_forward_mobile.cpp b/servers/rendering/renderer_rd/forward_mobile/scene_shader_forward_mobile.cpp index b5fb9fbc62..14b3b6d9aa 100644 --- a/servers/rendering/renderer_rd/forward_mobile/scene_shader_forward_mobile.cpp +++ b/servers/rendering/renderer_rd/forward_mobile/scene_shader_forward_mobile.cpp @@ -135,8 +135,7 @@ void SceneShaderForwardMobile::ShaderData::set_code(const String &p_code) { SceneShaderForwardMobile *shader_singleton = (SceneShaderForwardMobile *)SceneShaderForwardMobile::singleton; Error err = shader_singleton->compiler.compile(RS::SHADER_SPATIAL, code, &actions, path, gen_code); - - ERR_FAIL_COND(err != OK); + ERR_FAIL_COND_MSG(err != OK, "Shader compilation failed."); if (version.is_null()) { version = shader_singleton->shader.version_create(); @@ -318,7 +317,7 @@ void SceneShaderForwardMobile::ShaderData::set_code(const String &p_code) { } RID shader_variant = shader_singleton->shader.version_get_shader(version, k); - pipelines[i][j][k].setup(shader_variant, primitive_rd, raster_state, multisample_state, depth_stencil, blend_state, 0); + pipelines[i][j][k].setup(shader_variant, primitive_rd, raster_state, multisample_state, depth_stencil, blend_state, 0, singleton->default_specialization_constants); } } } @@ -402,7 +401,8 @@ RS::ShaderNativeSourceCode SceneShaderForwardMobile::ShaderData::get_native_sour return shader_singleton->shader.version_get_native_source_code(version); } -SceneShaderForwardMobile::ShaderData::ShaderData() { +SceneShaderForwardMobile::ShaderData::ShaderData() : + shader_list_element(this) { valid = false; uses_screen_texture = false; } @@ -418,6 +418,7 @@ SceneShaderForwardMobile::ShaderData::~ShaderData() { RendererStorageRD::ShaderData *SceneShaderForwardMobile::_create_shader_func() { ShaderData *shader_data = memnew(ShaderData); + singleton->shader_list.add(&shader_data->shader_list_element); return shader_data; } @@ -429,94 +430,14 @@ void SceneShaderForwardMobile::MaterialData::set_next_pass(RID p_pass) { next_pass = p_pass; } -void SceneShaderForwardMobile::MaterialData::update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) { +bool SceneShaderForwardMobile::MaterialData::update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) { SceneShaderForwardMobile *shader_singleton = (SceneShaderForwardMobile *)SceneShaderForwardMobile::singleton; - if ((uint32_t)ubo_data.size() != shader_data->ubo_size) { - p_uniform_dirty = true; - if (uniform_buffer.is_valid()) { - RD::get_singleton()->free(uniform_buffer); - uniform_buffer = RID(); - } - - ubo_data.resize(shader_data->ubo_size); - if (ubo_data.size()) { - uniform_buffer = RD::get_singleton()->uniform_buffer_create(ubo_data.size()); - memset(ubo_data.ptrw(), 0, ubo_data.size()); //clear - } - - //clear previous uniform set - if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - RD::get_singleton()->free(uniform_set); - uniform_set = RID(); - } - } - - //check whether buffer changed - if (p_uniform_dirty && ubo_data.size()) { - update_uniform_buffer(shader_data->uniforms, shader_data->ubo_offsets.ptr(), p_parameters, ubo_data.ptrw(), ubo_data.size(), false); - RD::get_singleton()->buffer_update(uniform_buffer, 0, ubo_data.size(), ubo_data.ptrw(), RD::BARRIER_MASK_RASTER); - } - - uint32_t tex_uniform_count = shader_data->texture_uniforms.size(); - - if ((uint32_t)texture_cache.size() != tex_uniform_count) { - texture_cache.resize(tex_uniform_count); - p_textures_dirty = true; - - //clear previous uniform set - if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - RD::get_singleton()->free(uniform_set); - uniform_set = RID(); - } - } - - if (p_textures_dirty && tex_uniform_count) { - update_textures(p_parameters, shader_data->default_texture_params, shader_data->texture_uniforms, texture_cache.ptrw(), true); - } - - if (shader_data->ubo_size == 0 && shader_data->texture_uniforms.size() == 0) { - // This material does not require an uniform set, so don't create it. - return; - } - - if (!p_textures_dirty && uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - //no reason to update uniform set, only UBO (or nothing) was needed to update - return; - } - - Vector<RD::Uniform> uniforms; - - { - if (shader_data->ubo_size) { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; - u.binding = 0; - u.ids.push_back(uniform_buffer); - uniforms.push_back(u); - } - - const RID *textures = texture_cache.ptrw(); - for (uint32_t i = 0; i < tex_uniform_count; i++) { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.binding = 1 + i; - u.ids.push_back(textures[i]); - uniforms.push_back(u); - } - } - - uniform_set = RD::get_singleton()->uniform_set_create(uniforms, shader_singleton->shader.version_get_shader(shader_data->version, 0), RenderForwardMobile::MATERIAL_UNIFORM_SET); + return update_parameters_uniform_set(p_parameters, p_uniform_dirty, p_textures_dirty, shader_data->uniforms, shader_data->ubo_offsets.ptr(), shader_data->texture_uniforms, shader_data->default_texture_params, shader_data->ubo_size, uniform_set, shader_singleton->shader.version_get_shader(shader_data->version, 0), RenderForwardMobile::MATERIAL_UNIFORM_SET, RD::BARRIER_MASK_RASTER); } SceneShaderForwardMobile::MaterialData::~MaterialData() { - if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - RD::get_singleton()->free(uniform_set); - } - - if (uniform_buffer.is_valid()) { - RD::get_singleton()->free(uniform_buffer); - } + free_parameters_uniform_set(uniform_set); } RendererStorageRD::MaterialData *SceneShaderForwardMobile::_create_material_func(ShaderData *p_shader) { @@ -621,7 +542,6 @@ void SceneShaderForwardMobile::init(RendererStorageRD *p_storage, const String p actions.renames["SSS_STRENGTH"] = "sss_strength"; actions.renames["SSS_TRANSMITTANCE_COLOR"] = "transmittance_color"; actions.renames["SSS_TRANSMITTANCE_DEPTH"] = "transmittance_depth"; - actions.renames["SSS_TRANSMITTANCE_CURVE"] = "transmittance_curve"; actions.renames["SSS_TRANSMITTANCE_BOOST"] = "transmittance_boost"; actions.renames["BACKLIGHT"] = "backlight"; actions.renames["AO"] = "ao"; @@ -745,6 +665,8 @@ void SceneShaderForwardMobile::init(RendererStorageRD *p_storage, const String p actions.global_buffer_array_variable = "global_variables.data"; actions.instance_uniform_index_variable = "draw_call.instance_uniforms_ofs"; + actions.apply_luminance_multiplier = true; // apply luminance multiplier to screen texture + compiler.initialize(actions); } @@ -752,29 +674,55 @@ void SceneShaderForwardMobile::init(RendererStorageRD *p_storage, const String p //default material and shader default_shader = storage->shader_allocate(); storage->shader_initialize(default_shader); - storage->shader_set_code(default_shader, "shader_type spatial; void vertex() { ROUGHNESS = 0.8; } void fragment() { ALBEDO=vec3(0.6); ROUGHNESS=0.8; METALLIC=0.2; } \n"); + storage->shader_set_code(default_shader, R"( +// Default 3D material shader (mobile). + +shader_type spatial; + +void vertex() { + ROUGHNESS = 0.8; +} + +void fragment() { + ALBEDO = vec3(0.6); + ROUGHNESS = 0.8; + METALLIC = 0.2; +} +)"); default_material = storage->material_allocate(); storage->material_initialize(default_material); storage->material_set_shader(default_material, default_shader); MaterialData *md = (MaterialData *)storage->material_get_data(default_material, RendererStorageRD::SHADER_TYPE_3D); default_shader_rd = shader.version_get_shader(md->shader_data->version, SHADER_VERSION_COLOR_PASS); + + default_material_shader_ptr = md->shader_data; + default_material_uniform_set = md->uniform_set; } { overdraw_material_shader = storage->shader_allocate(); storage->shader_initialize(overdraw_material_shader); - storage->shader_set_code(overdraw_material_shader, "shader_type spatial;\nrender_mode blend_add,unshaded;\n void fragment() { ALBEDO=vec3(0.4,0.8,0.8); ALPHA=0.2; }"); + // Use relatively low opacity so that more "layers" of overlapping objects can be distinguished. + storage->shader_set_code(overdraw_material_shader, R"( +// 3D editor Overdraw debug draw mode shader (mobile). + +shader_type spatial; + +render_mode blend_add, unshaded; + +void fragment() { + ALBEDO = vec3(0.4, 0.8, 0.8); + ALPHA = 0.1; +} +)"); overdraw_material = storage->material_allocate(); storage->material_initialize(overdraw_material); storage->material_set_shader(overdraw_material, overdraw_material_shader); - wireframe_material_shader = storage->shader_allocate(); - storage->shader_initialize(wireframe_material_shader); - storage->shader_set_code(wireframe_material_shader, "shader_type spatial;\nrender_mode wireframe,unshaded;\n void fragment() { ALBEDO=vec3(0.0,0.0,0.0); }"); - wireframe_material = storage->material_allocate(); - storage->material_initialize(wireframe_material); - storage->material_set_shader(wireframe_material, wireframe_material_shader); + MaterialData *md = (MaterialData *)storage->material_get_data(overdraw_material, RendererStorageRD::SHADER_TYPE_3D); + overdraw_material_shader_ptr = md->shader_data; + overdraw_material_uniform_set = md->uniform_set; } { @@ -798,15 +746,26 @@ void SceneShaderForwardMobile::init(RendererStorageRD *p_storage, const String p } } +void SceneShaderForwardMobile::set_default_specialization_constants(const Vector<RD::PipelineSpecializationConstant> &p_constants) { + default_specialization_constants = p_constants; + for (SelfList<ShaderData> *E = shader_list.first(); E; E = E->next()) { + for (int i = 0; i < ShaderData::CULL_VARIANT_MAX; i++) { + for (int j = 0; j < RS::PRIMITIVE_MAX; j++) { + for (int k = 0; k < SHADER_VERSION_MAX; k++) { + E->self()->pipelines[i][j][k].update_specialization_constants(default_specialization_constants); + } + } + } + } +} + SceneShaderForwardMobile::~SceneShaderForwardMobile() { RD::get_singleton()->free(default_vec4_xform_buffer); RD::get_singleton()->free(shadow_sampler); - storage->free(wireframe_material_shader); storage->free(overdraw_material_shader); storage->free(default_shader); - storage->free(wireframe_material); storage->free(overdraw_material); storage->free(default_material); } diff --git a/servers/rendering/renderer_rd/forward_mobile/scene_shader_forward_mobile.h b/servers/rendering/renderer_rd/forward_mobile/scene_shader_forward_mobile.h index f4f6ceeb1d..e1c10f0206 100644 --- a/servers/rendering/renderer_rd/forward_mobile/scene_shader_forward_mobile.h +++ b/servers/rendering/renderer_rd/forward_mobile/scene_shader_forward_mobile.h @@ -151,6 +151,8 @@ public: virtual Variant get_default_parameter(const StringName &p_parameter) const; virtual RS::ShaderNativeSourceCode get_native_source_code() const; + SelfList<ShaderData> shader_list_element; + ShaderData(); virtual ~ShaderData(); }; @@ -163,20 +165,19 @@ public: struct MaterialData : public RendererStorageRD::MaterialData { uint64_t last_frame; ShaderData *shader_data; - RID uniform_buffer; RID uniform_set; - Vector<RID> texture_cache; - Vector<uint8_t> ubo_data; uint64_t last_pass = 0; uint32_t index = 0; RID next_pass; uint8_t priority; virtual void set_render_priority(int p_priority); virtual void set_next_pass(RID p_pass); - virtual void update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty); + virtual bool update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty); virtual ~MaterialData(); }; + SelfList<ShaderData>::List shader_list; + RendererStorageRD::MaterialData *_create_material_func(ShaderData *p_shader); static RendererStorageRD::MaterialData *_create_material_funcs(RendererStorageRD::ShaderData *p_shader) { return static_cast<SceneShaderForwardMobile *>(singleton)->_create_material_func(static_cast<ShaderData *>(p_shader)); @@ -189,8 +190,6 @@ public: RID default_material; RID overdraw_material_shader; RID overdraw_material; - RID wireframe_material_shader; - RID wireframe_material; RID default_shader_rd; RID default_vec4_xform_buffer; @@ -198,10 +197,19 @@ public: RID shadow_sampler; + RID default_material_uniform_set; + ShaderData *default_material_shader_ptr = nullptr; + + RID overdraw_material_uniform_set; + ShaderData *overdraw_material_shader_ptr = nullptr; + SceneShaderForwardMobile(); ~SceneShaderForwardMobile(); + Vector<RD::PipelineSpecializationConstant> default_specialization_constants; + void init(RendererStorageRD *p_storage, const String p_defines); + void set_default_specialization_constants(const Vector<RD::PipelineSpecializationConstant> &p_constants); }; } // namespace RendererSceneRenderImplementation diff --git a/servers/rendering/renderer_rd/pipeline_cache_rd.cpp b/servers/rendering/renderer_rd/pipeline_cache_rd.cpp index b2b919c40e..aefe926cb0 100644 --- a/servers/rendering/renderer_rd/pipeline_cache_rd.cpp +++ b/servers/rendering/renderer_rd/pipeline_cache_rd.cpp @@ -31,25 +31,46 @@ #include "pipeline_cache_rd.h" #include "core/os/memory.h" -RID PipelineCacheRD::_generate_version(RD::VertexFormatID p_vertex_format_id, RD::FramebufferFormatID p_framebuffer_format_id, bool p_wireframe) { +RID PipelineCacheRD::_generate_version(RD::VertexFormatID p_vertex_format_id, RD::FramebufferFormatID p_framebuffer_format_id, bool p_wireframe, uint32_t p_render_pass, uint32_t p_bool_specializations) { RD::PipelineMultisampleState multisample_state_version = multisample_state; - multisample_state_version.sample_count = RD::get_singleton()->framebuffer_format_get_texture_samples(p_framebuffer_format_id); + multisample_state_version.sample_count = RD::get_singleton()->framebuffer_format_get_texture_samples(p_framebuffer_format_id, p_render_pass); RD::PipelineRasterizationState raster_state_version = rasterization_state; raster_state_version.wireframe = p_wireframe; - RID pipeline = RD::get_singleton()->render_pipeline_create(shader, p_framebuffer_format_id, p_vertex_format_id, render_primitive, raster_state_version, multisample_state_version, depth_stencil_state, blend_state, dynamic_state_flags); + Vector<RD::PipelineSpecializationConstant> specialization_constants = base_specialization_constants; + + uint32_t bool_index = 0; + uint32_t bool_specializations = p_bool_specializations; + while (bool_specializations) { + if (bool_specializations & (1 << bool_index)) { + RD::PipelineSpecializationConstant sc; + sc.bool_value = true; + sc.constant_id = bool_index; + sc.type = RD::PIPELINE_SPECIALIZATION_CONSTANT_TYPE_BOOL; + specialization_constants.push_back(sc); + bool_specializations &= ~(1 << bool_index); + } + bool_index++; + } + + RID pipeline = RD::get_singleton()->render_pipeline_create(shader, p_framebuffer_format_id, p_vertex_format_id, render_primitive, raster_state_version, multisample_state_version, depth_stencil_state, blend_state, dynamic_state_flags, p_render_pass, specialization_constants); ERR_FAIL_COND_V(pipeline.is_null(), RID()); versions = (Version *)memrealloc(versions, sizeof(Version) * (version_count + 1)); versions[version_count].framebuffer_id = p_framebuffer_format_id; versions[version_count].vertex_id = p_vertex_format_id; versions[version_count].wireframe = p_wireframe; versions[version_count].pipeline = pipeline; + versions[version_count].render_pass = p_render_pass; + versions[version_count].bool_specializations = p_bool_specializations; version_count++; return pipeline; } void PipelineCacheRD::_clear() { +#ifndef _MSC_VER +#warning Clear should probably recompile all the variants already compiled instead to avoid stalls? needs discussion +#endif if (versions) { for (uint32_t i = 0; i < version_count; i++) { //shader may be gone, so this may not be valid @@ -63,7 +84,7 @@ void PipelineCacheRD::_clear() { } } -void PipelineCacheRD::setup(RID p_shader, RD::RenderPrimitive p_primitive, const RD::PipelineRasterizationState &p_rasterization_state, RD::PipelineMultisampleState p_multisample, const RD::PipelineDepthStencilState &p_depth_stencil_state, const RD::PipelineColorBlendState &p_blend_state, int p_dynamic_state_flags) { +void PipelineCacheRD::setup(RID p_shader, RD::RenderPrimitive p_primitive, const RD::PipelineRasterizationState &p_rasterization_state, RD::PipelineMultisampleState p_multisample, const RD::PipelineDepthStencilState &p_depth_stencil_state, const RD::PipelineColorBlendState &p_blend_state, int p_dynamic_state_flags, const Vector<RD::PipelineSpecializationConstant> &p_base_specialization_constants) { ERR_FAIL_COND(p_shader.is_null()); _clear(); shader = p_shader; @@ -74,6 +95,11 @@ void PipelineCacheRD::setup(RID p_shader, RD::RenderPrimitive p_primitive, const depth_stencil_state = p_depth_stencil_state; blend_state = p_blend_state; dynamic_state_flags = p_dynamic_state_flags; + base_specialization_constants = p_base_specialization_constants; +} +void PipelineCacheRD::update_specialization_constants(const Vector<RD::PipelineSpecializationConstant> &p_base_specialization_constants) { + base_specialization_constants = p_base_specialization_constants; + _clear(); } void PipelineCacheRD::update_shader(RID p_shader) { diff --git a/servers/rendering/renderer_rd/pipeline_cache_rd.h b/servers/rendering/renderer_rd/pipeline_cache_rd.h index b1c8f21ecc..e52f47fa47 100644 --- a/servers/rendering/renderer_rd/pipeline_cache_rd.h +++ b/servers/rendering/renderer_rd/pipeline_cache_rd.h @@ -46,26 +46,30 @@ class PipelineCacheRD { RD::PipelineDepthStencilState depth_stencil_state; RD::PipelineColorBlendState blend_state; int dynamic_state_flags; + Vector<RD::PipelineSpecializationConstant> base_specialization_constants; struct Version { RD::VertexFormatID vertex_id; RD::FramebufferFormatID framebuffer_id; + uint32_t render_pass; bool wireframe; + uint32_t bool_specializations; RID pipeline; }; Version *versions; uint32_t version_count; - RID _generate_version(RD::VertexFormatID p_vertex_format_id, RD::FramebufferFormatID p_framebuffer_format_id, bool p_wireframe); + RID _generate_version(RD::VertexFormatID p_vertex_format_id, RD::FramebufferFormatID p_framebuffer_format_id, bool p_wireframe, uint32_t p_render_pass, uint32_t p_bool_specializations = 0); void _clear(); public: - void setup(RID p_shader, RD::RenderPrimitive p_primitive, const RD::PipelineRasterizationState &p_rasterization_state, RD::PipelineMultisampleState p_multisample, const RD::PipelineDepthStencilState &p_depth_stencil_state, const RD::PipelineColorBlendState &p_blend_state, int p_dynamic_state_flags = 0); + void setup(RID p_shader, RD::RenderPrimitive p_primitive, const RD::PipelineRasterizationState &p_rasterization_state, RD::PipelineMultisampleState p_multisample, const RD::PipelineDepthStencilState &p_depth_stencil_state, const RD::PipelineColorBlendState &p_blend_state, int p_dynamic_state_flags = 0, const Vector<RD::PipelineSpecializationConstant> &p_base_specialization_constants = Vector<RD::PipelineSpecializationConstant>()); + void update_specialization_constants(const Vector<RD::PipelineSpecializationConstant> &p_base_specialization_constants); void update_shader(RID p_shader); - _FORCE_INLINE_ RID get_render_pipeline(RD::VertexFormatID p_vertex_format_id, RD::FramebufferFormatID p_framebuffer_format_id, bool p_wireframe = false) { + _FORCE_INLINE_ RID get_render_pipeline(RD::VertexFormatID p_vertex_format_id, RD::FramebufferFormatID p_framebuffer_format_id, bool p_wireframe = false, uint32_t p_render_pass = 0, uint32_t p_bool_specializations = 0) { #ifdef DEBUG_ENABLED ERR_FAIL_COND_V_MSG(shader.is_null(), RID(), "Attempted to use an unused shader variant (shader is null),"); @@ -74,13 +78,13 @@ public: spin_lock.lock(); RID result; for (uint32_t i = 0; i < version_count; i++) { - if (versions[i].vertex_id == p_vertex_format_id && versions[i].framebuffer_id == p_framebuffer_format_id && versions[i].wireframe == p_wireframe) { + if (versions[i].vertex_id == p_vertex_format_id && versions[i].framebuffer_id == p_framebuffer_format_id && versions[i].wireframe == p_wireframe && versions[i].render_pass == p_render_pass && versions[i].bool_specializations == p_bool_specializations) { result = versions[i].pipeline; spin_lock.unlock(); return result; } } - result = _generate_version(p_vertex_format_id, p_framebuffer_format_id, p_wireframe); + result = _generate_version(p_vertex_format_id, p_framebuffer_format_id, p_wireframe, p_render_pass, p_bool_specializations); spin_lock.unlock(); return result; } diff --git a/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp b/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp index 9d325fe69b..647c348d9f 100644 --- a/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp +++ b/servers/rendering/renderer_rd/renderer_canvas_render_rd.cpp @@ -408,6 +408,7 @@ void RendererCanvasRenderRD::_render_item(RD::DrawListID p_draw_list, RID p_rend PushConstant push_constant; Transform2D base_transform = p_canvas_transform_inverse * p_item->final_transform; + Transform2D draw_transform; _update_transform_2d_to_mat2x3(base_transform, push_constant.world); Color base_color = p_item->final_modulate; @@ -464,14 +465,25 @@ void RendererCanvasRenderRD::_render_item(RD::DrawListID p_draw_list, RID p_rend RID last_texture; Size2 texpixel_size; + bool skipping = false; + const Item::Command *c = p_item->commands; while (c) { + if (skipping && c->type != Item::Command::TYPE_ANIMATION_SLICE) { + c = c->next; + continue; + } + push_constant.flags = base_flags | (push_constant.flags & (FLAGS_DEFAULT_NORMAL_MAP_USED | FLAGS_DEFAULT_SPECULAR_MAP_USED)); //reset on each command for sanity, keep canvastexture binding config switch (c->type) { case Item::Command::TYPE_RECT: { const Item::CommandRect *rect = static_cast<const Item::CommandRect *>(c); + if (rect->flags & CANVAS_RECT_TILE) { + current_repeat = RenderingServer::CanvasItemTextureRepeat::CANVAS_ITEM_TEXTURE_REPEAT_ENABLED; + } + //bind pipeline { RID pipeline = pipeline_variants->variants[light_mode][PIPELINE_VARIANT_QUAD].get_render_pipeline(RD::INVALID_ID, p_framebuffer_format); @@ -529,6 +541,14 @@ void RendererCanvasRenderRD::_render_item(RD::DrawListID p_draw_list, RID p_rend src_rect = Rect2(0, 0, 1, 1); } + if (rect->flags & CANVAS_RECT_MSDF) { + push_constant.flags |= FLAGS_USE_MSDF; + push_constant.msdf[0] = rect->px_range; // Pixel range. + push_constant.msdf[1] = rect->outline; // Outline size. + push_constant.msdf[2] = 0.f; // Reserved. + push_constant.msdf[3] = 0.f; // Reserved. + } + push_constant.modulation[0] = rect->modulate.r * base_color.r; push_constant.modulation[1] = rect->modulate.g * base_color.g; push_constant.modulation[2] = rect->modulate.b * base_color.b; @@ -612,7 +632,7 @@ void RendererCanvasRenderRD::_render_item(RD::DrawListID p_draw_list, RID p_rend RD::get_singleton()->draw_list_bind_index_array(p_draw_list, shader.quad_index_array); RD::get_singleton()->draw_list_draw(p_draw_list, true); - //restore if overrided + // Restore if overridden. push_constant.color_texture_pixel_size[0] = texpixel_size.x; push_constant.color_texture_pixel_size[1] = texpixel_size.y; @@ -724,7 +744,7 @@ void RendererCanvasRenderRD::_render_item(RD::DrawListID p_draw_list, RID p_rend mesh_instance = m->mesh_instance; texture = m->texture; modulate = m->modulate; - _update_transform_2d_to_mat2x3(base_transform * m->transform, push_constant.world); + _update_transform_2d_to_mat2x3(base_transform * draw_transform * m->transform, push_constant.world); } else if (c->type == Item::Command::TYPE_MULTIMESH) { const Item::CommandMultiMesh *mm = static_cast<const Item::CommandMultiMesh *>(c); RID multimesh = mm->multimesh; @@ -857,10 +877,10 @@ void RendererCanvasRenderRD::_render_item(RD::DrawListID p_draw_list, RID p_rend for (int j = 0; j < 6; j++) { push_constant.world[j] = world_backup[j]; } - } break; case Item::Command::TYPE_TRANSFORM: { const Item::CommandTransform *transform = static_cast<const Item::CommandTransform *>(c); + draw_transform = transform->xform; _update_transform_2d_to_mat2x3(base_transform * transform->xform, push_constant.world); } break; @@ -879,6 +899,14 @@ void RendererCanvasRenderRD::_render_item(RD::DrawListID p_draw_list, RID p_rend } } break; + case Item::Command::TYPE_ANIMATION_SLICE: { + const Item::CommandAnimationSlice *as = static_cast<const Item::CommandAnimationSlice *>(c); + double current_time = RendererCompositorRD::singleton->get_total_time(); + double local_time = Math::fposmod(current_time - as->offset, as->animation_length); + skipping = !(local_time >= as->slice_begin && local_time < as->slice_end); + + RenderingServerDefault::redraw_request(); // animation visible means redraw request + } break; } c = c->next; @@ -1073,7 +1101,8 @@ void RendererCanvasRenderRD::_render_items(RID p_to_render_target, int p_item_co if (material_data) { if (material_data->shader_data->version.is_valid() && material_data->shader_data->valid) { pipeline_variants = &material_data->shader_data->pipeline_variants; - if (material_data->uniform_set.is_valid()) { + // Update uniform set. + if (material_data->uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(material_data->uniform_set)) { // Material may not have a uniform set. RD::get_singleton()->draw_list_bind_uniform_set(draw_list, material_data->uniform_set, MATERIAL_UNIFORM_SET); } } else { @@ -1949,8 +1978,7 @@ void RendererCanvasRenderRD::ShaderData::set_code(const String &p_code) { RendererCanvasRenderRD *canvas_singleton = (RendererCanvasRenderRD *)RendererCanvasRender::singleton; Error err = canvas_singleton->shader.compiler.compile(RS::SHADER_CANVAS_ITEM, code, &actions, path, gen_code); - - ERR_FAIL_COND(err != OK); + ERR_FAIL_COND_MSG(err != OK, "Shader compilation failed."); if (version.is_null()) { version = canvas_singleton->shader.canvas_shader.version_create(); @@ -2185,94 +2213,14 @@ RendererStorageRD::ShaderData *RendererCanvasRenderRD::_create_shader_func() { return shader_data; } -void RendererCanvasRenderRD::MaterialData::update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) { +bool RendererCanvasRenderRD::MaterialData::update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) { RendererCanvasRenderRD *canvas_singleton = (RendererCanvasRenderRD *)RendererCanvasRender::singleton; - if ((uint32_t)ubo_data.size() != shader_data->ubo_size) { - p_uniform_dirty = true; - if (uniform_buffer.is_valid()) { - RD::get_singleton()->free(uniform_buffer); - uniform_buffer = RID(); - } - - ubo_data.resize(shader_data->ubo_size); - if (ubo_data.size()) { - uniform_buffer = RD::get_singleton()->uniform_buffer_create(ubo_data.size()); - memset(ubo_data.ptrw(), 0, ubo_data.size()); //clear - } - - //clear previous uniform set - if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - RD::get_singleton()->free(uniform_set); - uniform_set = RID(); - } - } - - //check whether buffer changed - if (p_uniform_dirty && ubo_data.size()) { - update_uniform_buffer(shader_data->uniforms, shader_data->ubo_offsets.ptr(), p_parameters, ubo_data.ptrw(), ubo_data.size(), false); - RD::get_singleton()->buffer_update(uniform_buffer, 0, ubo_data.size(), ubo_data.ptrw()); - } - - uint32_t tex_uniform_count = shader_data->texture_uniforms.size(); - - if ((uint32_t)texture_cache.size() != tex_uniform_count) { - texture_cache.resize(tex_uniform_count); - p_textures_dirty = true; - - //clear previous uniform set - if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - RD::get_singleton()->free(uniform_set); - uniform_set = RID(); - } - } - - if (p_textures_dirty && tex_uniform_count) { - update_textures(p_parameters, shader_data->default_texture_params, shader_data->texture_uniforms, texture_cache.ptrw(), false); - } - - if (shader_data->ubo_size == 0) { - // This material does not require an uniform set, so don't create it. - return; - } - - if (!p_textures_dirty && uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - //no reason to update uniform set, only UBO (or nothing) was needed to update - return; - } - - Vector<RD::Uniform> uniforms; - - { - if (shader_data->ubo_size) { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; - u.binding = 0; - u.ids.push_back(uniform_buffer); - uniforms.push_back(u); - } - - const RID *textures = texture_cache.ptrw(); - for (uint32_t i = 0; i < tex_uniform_count; i++) { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.binding = 1 + i; - u.ids.push_back(textures[i]); - uniforms.push_back(u); - } - } - - uniform_set = RD::get_singleton()->uniform_set_create(uniforms, canvas_singleton->shader.canvas_shader.version_get_shader(shader_data->version, 0), MATERIAL_UNIFORM_SET); + return update_parameters_uniform_set(p_parameters, p_uniform_dirty, p_textures_dirty, shader_data->uniforms, shader_data->ubo_offsets.ptr(), shader_data->texture_uniforms, shader_data->default_texture_params, shader_data->ubo_size, uniform_set, canvas_singleton->shader.canvas_shader.version_get_shader(shader_data->version, 0), MATERIAL_UNIFORM_SET); } RendererCanvasRenderRD::MaterialData::~MaterialData() { - if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - RD::get_singleton()->free(uniform_set); - } - - if (uniform_buffer.is_valid()) { - RD::get_singleton()->free(uniform_buffer); - } + free_parameters_uniform_set(uniform_set); } RendererStorageRD::MaterialData *RendererCanvasRenderRD::_create_material_func(ShaderData *p_shader) { @@ -2634,8 +2582,21 @@ RendererCanvasRenderRD::RendererCanvasRenderRD(RendererStorageRD *p_storage) { default_canvas_group_shader = storage->shader_allocate(); storage->shader_initialize(default_canvas_group_shader); - storage->shader_set_code(default_canvas_group_shader, "shader_type canvas_item; \nvoid fragment() {\n\tvec4 c = textureLod(SCREEN_TEXTURE,SCREEN_UV,0.0); if (c.a > 0.0001) c.rgb/=c.a; COLOR *= c; \n}\n"); + storage->shader_set_code(default_canvas_group_shader, R"( +// Default CanvasGroup shader. + +shader_type canvas_item; + +void fragment() { + vec4 c = textureLod(SCREEN_TEXTURE, SCREEN_UV, 0.0); + if (c.a > 0.0001) { + c.rgb /= c.a; + } + + COLOR *= c; +} +)"); default_canvas_group_material = storage->material_allocate(); storage->material_initialize(default_canvas_group_material); diff --git a/servers/rendering/renderer_rd/renderer_canvas_render_rd.h b/servers/rendering/renderer_rd/renderer_canvas_render_rd.h index 8c1376e2dc..ec7d7e2854 100644 --- a/servers/rendering/renderer_rd/renderer_canvas_render_rd.h +++ b/servers/rendering/renderer_rd/renderer_canvas_render_rd.h @@ -84,8 +84,9 @@ class RendererCanvasRenderRD : public RendererCanvasRender { FLAGS_LIGHT_COUNT_SHIFT = 20, FLAGS_DEFAULT_NORMAL_MAP_USED = (1 << 26), - FLAGS_DEFAULT_SPECULAR_MAP_USED = (1 << 27) + FLAGS_DEFAULT_SPECULAR_MAP_USED = (1 << 27), + FLAGS_USE_MSDF = (1 << 28), }; enum { @@ -201,14 +202,11 @@ class RendererCanvasRenderRD : public RendererCanvasRender { struct MaterialData : public RendererStorageRD::MaterialData { uint64_t last_frame; ShaderData *shader_data; - RID uniform_buffer; RID uniform_set; - Vector<RID> texture_cache; - Vector<uint8_t> ubo_data; virtual void set_render_priority(int p_priority) {} virtual void set_next_pass(RID p_pass) {} - virtual void update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty); + virtual bool update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty); virtual ~MaterialData(); }; @@ -391,7 +389,10 @@ class RendererCanvasRenderRD : public RendererCanvasRender { //rect struct { float modulation[4]; - float ninepatch_margins[4]; + union { + float msdf[4]; + float ninepatch_margins[4]; + }; float dst_rect[4]; float src_rect[4]; float pad[2]; @@ -457,8 +458,6 @@ public: void canvas_debug_viewport_shadows(Light *p_lights_with_shadow) {} - void draw_window_margins(int *p_margins, RID *p_margin_textures) {} - virtual void set_shadow_texture_size(int p_size); void set_time(double p_time); diff --git a/servers/rendering/renderer_rd/renderer_compositor_rd.cpp b/servers/rendering/renderer_rd/renderer_compositor_rd.cpp index f9ac7c8fa3..c53c202bab 100644 --- a/servers/rendering/renderer_rd/renderer_compositor_rd.cpp +++ b/servers/rendering/renderer_rd/renderer_compositor_rd.cpp @@ -46,6 +46,8 @@ void RendererCompositorRD::blit_render_targets_to_screen(DisplayServer::WindowID RID rd_texture = storage->texture_get_rd_texture(texture); ERR_CONTINUE(rd_texture.is_null()); + // TODO if keep_3d_linear was set when rendering to this render target we need to add a linear->sRGB conversion in. + if (!render_target_descriptors.has(rd_texture) || !RD::get_singleton()->uniform_set_is_valid(render_target_descriptors[rd_texture])) { Vector<RD::Uniform> uniforms; RD::Uniform u; @@ -60,16 +62,19 @@ void RendererCompositorRD::blit_render_targets_to_screen(DisplayServer::WindowID } Size2 screen_size(RD::get_singleton()->screen_get_width(p_screen), RD::get_singleton()->screen_get_height(p_screen)); - BlitMode mode = p_render_targets[i].lens_distortion.apply ? BLIT_MODE_LENS : p_render_targets[i].multi_view.use_layer ? BLIT_MODE_USE_LAYER : - BLIT_MODE_NORMAL; + BlitMode mode = p_render_targets[i].lens_distortion.apply ? BLIT_MODE_LENS : (p_render_targets[i].multi_view.use_layer ? BLIT_MODE_USE_LAYER : BLIT_MODE_NORMAL); RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, blit.pipelines[mode]); RD::get_singleton()->draw_list_bind_index_array(draw_list, blit.array); RD::get_singleton()->draw_list_bind_uniform_set(draw_list, render_target_descriptors[rd_texture], 0); - blit.push_constant.rect[0] = p_render_targets[i].rect.position.x / screen_size.width; - blit.push_constant.rect[1] = p_render_targets[i].rect.position.y / screen_size.height; - blit.push_constant.rect[2] = p_render_targets[i].rect.size.width / screen_size.width; - blit.push_constant.rect[3] = p_render_targets[i].rect.size.height / screen_size.height; + blit.push_constant.src_rect[0] = p_render_targets[i].src_rect.position.x; + blit.push_constant.src_rect[1] = p_render_targets[i].src_rect.position.y; + blit.push_constant.src_rect[2] = p_render_targets[i].src_rect.size.width; + blit.push_constant.src_rect[3] = p_render_targets[i].src_rect.size.height; + blit.push_constant.dst_rect[0] = p_render_targets[i].dst_rect.position.x / screen_size.width; + blit.push_constant.dst_rect[1] = p_render_targets[i].dst_rect.position.y / screen_size.height; + blit.push_constant.dst_rect[2] = p_render_targets[i].dst_rect.size.width / screen_size.width; + blit.push_constant.dst_rect[3] = p_render_targets[i].dst_rect.size.height / screen_size.height; blit.push_constant.layer = p_render_targets[i].multi_view.layer; blit.push_constant.eye_center[0] = p_render_targets[i].lens_distortion.eye_center.x; blit.push_constant.eye_center[1] = p_render_targets[i].lens_distortion.eye_center.y; @@ -111,13 +116,14 @@ void RendererCompositorRD::initialize() { blit_modes.push_back("\n"); blit_modes.push_back("\n#define USE_LAYER\n"); blit_modes.push_back("\n#define USE_LAYER\n#define APPLY_LENS_DISTORTION\n"); + blit_modes.push_back("\n"); blit.shader.initialize(blit_modes); blit.shader_version = blit.shader.version_create(); for (int i = 0; i < BLIT_MODE_MAX; i++) { - blit.pipelines[i] = RD::get_singleton()->render_pipeline_create(blit.shader.version_get_shader(blit.shader_version, i), RD::get_singleton()->screen_get_framebuffer_format(), RD::INVALID_ID, RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), RenderingDevice::PipelineColorBlendState::create_disabled(), 0); + blit.pipelines[i] = RD::get_singleton()->render_pipeline_create(blit.shader.version_get_shader(blit.shader_version, i), RD::get_singleton()->screen_get_framebuffer_format(), RD::INVALID_ID, RD::RENDER_PRIMITIVE_TRIANGLES, RD::PipelineRasterizationState(), RD::PipelineMultisampleState(), RD::PipelineDepthStencilState(), i == BLIT_MODE_NORMAL_ALPHA ? RenderingDevice::PipelineColorBlendState::create_blend() : RenderingDevice::PipelineColorBlendState::create_disabled(), 0); } //create index array for copy shader @@ -153,6 +159,82 @@ void RendererCompositorRD::finalize() { RD::get_singleton()->free(blit.sampler); } +void RendererCompositorRD::set_boot_image(const Ref<Image> &p_image, const Color &p_color, bool p_scale, bool p_use_filter) { + RD::get_singleton()->prepare_screen_for_drawing(); + + RID texture = storage->texture_allocate(); + storage->texture_2d_initialize(texture, p_image); + RID rd_texture = storage->texture_get_rd_texture(texture); + + RID uset; + { + Vector<RD::Uniform> uniforms; + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_SAMPLER_WITH_TEXTURE; + u.binding = 0; + u.ids.push_back(blit.sampler); + u.ids.push_back(rd_texture); + uniforms.push_back(u); + uset = RD::get_singleton()->uniform_set_create(uniforms, blit.shader.version_get_shader(blit.shader_version, BLIT_MODE_NORMAL), 0); + } + + Size2 window_size = DisplayServer::get_singleton()->window_get_size(); + + Rect2 imgrect(0, 0, p_image->get_width(), p_image->get_height()); + Rect2 screenrect; + if (p_scale) { + if (window_size.width > window_size.height) { + //scale horizontally + screenrect.size.y = window_size.height; + screenrect.size.x = imgrect.size.x * window_size.height / imgrect.size.y; + screenrect.position.x = (window_size.width - screenrect.size.x) / 2; + + } else { + //scale vertically + screenrect.size.x = window_size.width; + screenrect.size.y = imgrect.size.y * window_size.width / imgrect.size.x; + screenrect.position.y = (window_size.height - screenrect.size.y) / 2; + } + } else { + screenrect = imgrect; + screenrect.position += ((Size2(window_size.width, window_size.height) - screenrect.size) / 2.0).floor(); + } + + screenrect.position /= window_size; + screenrect.size /= window_size; + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin_for_screen(DisplayServer::MAIN_WINDOW_ID, p_color); + + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, blit.pipelines[BLIT_MODE_NORMAL_ALPHA]); + RD::get_singleton()->draw_list_bind_index_array(draw_list, blit.array); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, uset, 0); + + blit.push_constant.src_rect[0] = 0.0; + blit.push_constant.src_rect[1] = 0.0; + blit.push_constant.src_rect[2] = 1.0; + blit.push_constant.src_rect[3] = 1.0; + blit.push_constant.dst_rect[0] = screenrect.position.x; + blit.push_constant.dst_rect[1] = screenrect.position.y; + blit.push_constant.dst_rect[2] = screenrect.size.width; + blit.push_constant.dst_rect[3] = screenrect.size.height; + blit.push_constant.layer = 0; + blit.push_constant.eye_center[0] = 0; + blit.push_constant.eye_center[1] = 0; + blit.push_constant.k1 = 0; + blit.push_constant.k2 = 0; + blit.push_constant.upscale = 1.0; + blit.push_constant.aspect_ratio = 1.0; + + RD::get_singleton()->draw_list_set_push_constant(draw_list, &blit.push_constant, sizeof(BlitPushConstant)); + RD::get_singleton()->draw_list_draw(draw_list, true); + + RD::get_singleton()->draw_list_end(); + + RD::get_singleton()->swap_buffers(); + + storage->free(texture); +} + RendererCompositorRD *RendererCompositorRD::singleton = nullptr; RendererCompositorRD::RendererCompositorRD() { @@ -208,6 +290,11 @@ RendererCompositorRD::RendererCompositorRD() { // default to our high end renderer scene = memnew(RendererSceneRenderImplementation::RenderForwardClustered(storage)); } + + scene->init(); + + // now we're ready to create our effects, + storage->init_effects(!scene->_render_buffers_can_be_storage()); } RendererCompositorRD::~RendererCompositorRD() { diff --git a/servers/rendering/renderer_rd/renderer_compositor_rd.h b/servers/rendering/renderer_rd/renderer_compositor_rd.h index 7a78322051..0230c46800 100644 --- a/servers/rendering/renderer_rd/renderer_compositor_rd.h +++ b/servers/rendering/renderer_rd/renderer_compositor_rd.h @@ -50,11 +50,13 @@ protected: BLIT_MODE_NORMAL, BLIT_MODE_USE_LAYER, BLIT_MODE_LENS, + BLIT_MODE_NORMAL_ALPHA, BLIT_MODE_MAX }; struct BlitPushConstant { - float rect[4]; + float src_rect[4]; + float dst_rect[4]; float eye_center[2]; float k1; @@ -79,7 +81,7 @@ protected: Map<RID, RID> render_target_descriptors; double time; - float delta; + double delta; static uint64_t frame; @@ -88,7 +90,7 @@ public: RendererCanvasRender *get_canvas() { return canvas; } RendererSceneRender *get_scene() { return scene; } - void set_boot_image(const Ref<Image> &p_image, const Color &p_color, bool p_scale, bool p_use_filter) {} + void set_boot_image(const Ref<Image> &p_image, const Color &p_color, bool p_scale, bool p_use_filter); void initialize(); void begin_frame(double frame_step); @@ -99,7 +101,7 @@ public: void finalize(); _ALWAYS_INLINE_ uint64_t get_frame_number() const { return frame; } - _ALWAYS_INLINE_ float get_frame_delta_time() const { return delta; } + _ALWAYS_INLINE_ double get_frame_delta_time() const { return delta; } _ALWAYS_INLINE_ double get_total_time() const { return time; } static Error is_viable() { diff --git a/servers/rendering/renderer_rd/renderer_scene_gi_rd.cpp b/servers/rendering/renderer_rd/renderer_scene_gi_rd.cpp index 43a4058ab6..098e2a5c87 100644 --- a/servers/rendering/renderer_rd/renderer_scene_gi_rd.cpp +++ b/servers/rendering/renderer_rd/renderer_scene_gi_rd.cpp @@ -1862,7 +1862,7 @@ void RendererSceneGIRD::SDFGI::render_region(RID p_render_buffers, int p_region, #if 0 Vector<uint8_t> data = RD::get_singleton()->texture_get_data(cascades[cascade].sdf, 0); Ref<Image> img; - img.instance(); + img.instantiate(); for (uint32_t i = 0; i < cascade_size; i++) { Vector<uint8_t> subarr = data.subarray(128 * 128 * i, 128 * 128 * (i + 1) - 1); img->create(cascade_size, cascade_size, false, Image::FORMAT_L8, subarr); @@ -1875,7 +1875,7 @@ void RendererSceneGIRD::SDFGI::render_region(RID p_render_buffers, int p_region, #if 0 Vector<uint8_t> data = RD::get_singleton()->texture_get_data(render_albedo, 0); Ref<Image> img; - img.instance(); + img.instantiate(); for (uint32_t i = 0; i < cascade_size; i++) { Vector<uint8_t> subarr = data.subarray(128 * 128 * i * 2, 128 * 128 * (i + 1) * 2 - 1); img->createcascade_size, cascade_size, false, Image::FORMAT_RGB565, subarr); @@ -2812,8 +2812,6 @@ void RendererSceneGIRD::init(RendererStorageRD *p_storage, RendererSceneSkyRD *p { //kinda complicated to compute the amount of slots, we try to use as many as we can - voxel_gi_max_lights = 32; - voxel_gi_lights = memnew_arr(VoxelGILight, voxel_gi_max_lights); voxel_gi_lights_uniform = RD::get_singleton()->uniform_buffer_create(voxel_gi_max_lights * sizeof(VoxelGILight)); voxel_gi_quality = RS::VoxelGIQuality(CLAMP(int(GLOBAL_GET("rendering/global_illumination/voxel_gi/quality")), 0, 1)); @@ -3009,7 +3007,9 @@ void RendererSceneGIRD::free() { sdfgi_shader.integrate.version_free(sdfgi_shader.integrate_shader); sdfgi_shader.preprocess.version_free(sdfgi_shader.preprocess_shader); - memdelete_arr(voxel_gi_lights); + if (voxel_gi_lights) { + memdelete_arr(voxel_gi_lights); + } } RendererSceneGIRD::SDFGI *RendererSceneGIRD::create_sdfgi(RendererSceneEnvironmentRD *p_env, const Vector3 &p_world_position, uint32_t p_requested_history_size) { @@ -3029,8 +3029,6 @@ void RendererSceneGIRD::setup_voxel_gi_instances(RID p_render_buffers, const Tra RID voxel_gi_buffer = p_scene_render->render_buffers_get_voxel_gi_buffer(p_render_buffers); - RD::get_singleton()->draw_command_begin_label("VoxelGIs Setup"); - VoxelGIData voxel_gi_data[MAX_VOXEL_GI_INSTANCES]; bool voxel_gi_instances_changed = false; @@ -3078,9 +3076,6 @@ void RendererSceneGIRD::setup_voxel_gi_instances(RID p_render_buffers, const Tra gipd.bias = storage->voxel_gi_get_bias(base_probe); gipd.normal_bias = storage->voxel_gi_get_normal_bias(base_probe); gipd.blend_ambient = !storage->voxel_gi_is_interior(base_probe); - gipd.anisotropy_strength = 0; - gipd.ao = storage->voxel_gi_get_ao(base_probe); - gipd.ao_size = Math::pow(storage->voxel_gi_get_ao_size(base_probe), 4.0f); gipd.mipmaps = gipi->mipmaps.size(); } @@ -3113,10 +3108,12 @@ void RendererSceneGIRD::setup_voxel_gi_instances(RID p_render_buffers, const Tra } if (p_voxel_gi_instances.size() > 0) { + RD::get_singleton()->draw_command_begin_label("VoxelGIs Setup"); + RD::get_singleton()->buffer_update(voxel_gi_buffer, 0, sizeof(VoxelGIData) * MIN((uint64_t)MAX_VOXEL_GI_INSTANCES, p_voxel_gi_instances.size()), voxel_gi_data, RD::BARRIER_MASK_COMPUTE); - } - RD::get_singleton()->draw_command_end_label(); + RD::get_singleton()->draw_command_end_label(); + } } void RendererSceneGIRD::process_gi(RID p_render_buffers, RID p_normal_roughness_buffer, RID p_voxel_gi_buffer, RID p_environment, const CameraMatrix &p_projection, const Transform3D &p_transform, const PagedArray<RID> &p_voxel_gi_instances, RendererSceneRenderRD *p_scene_render) { @@ -3349,6 +3346,7 @@ void RendererSceneGIRD::process_gi(RID p_render_buffers, RID p_normal_roughness_ } else { mode = (use_sdfgi && use_voxel_gi_instances) ? MODE_COMBINED : (use_sdfgi ? MODE_SDFGI : MODE_VOXEL_GI); } + RD::ComputeListID compute_list = RD::get_singleton()->compute_list_begin(true); RD::get_singleton()->compute_list_bind_compute_pipeline(compute_list, pipelines[mode]); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, rb->gi.uniform_set, 0); diff --git a/servers/rendering/renderer_rd/renderer_scene_gi_rd.h b/servers/rendering/renderer_rd/renderer_scene_gi_rd.h index 45fc7b3951..0b4622646f 100644 --- a/servers/rendering/renderer_rd/renderer_scene_gi_rd.h +++ b/servers/rendering/renderer_rd/renderer_scene_gi_rd.h @@ -110,8 +110,8 @@ private: float pad[3]; }; - VoxelGILight *voxel_gi_lights; - uint32_t voxel_gi_max_lights; + VoxelGILight *voxel_gi_lights = nullptr; + uint32_t voxel_gi_max_lights = 32; RID voxel_gi_lights_uniform; enum { @@ -611,9 +611,9 @@ public: uint32_t blend_ambient; uint32_t texture_slot; - float anisotropy_strength; - float ao; - float ao_size; + uint32_t pad0; + uint32_t pad1; + uint32_t pad2; uint32_t mipmaps; }; diff --git a/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp b/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp index 6e755bac4e..fa66ed85a9 100644 --- a/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp +++ b/servers/rendering/renderer_rd/renderer_scene_render_rd.cpp @@ -486,7 +486,7 @@ Ref<Image> RendererSceneRenderRD::environment_bake_panorama(RID p_env, bool p_ba color.b *= env->bg_energy; Ref<Image> ret; - ret.instance(); + ret.instantiate(); ret->create(p_size.width, p_size.height, false, Image::FORMAT_RGBAF); for (int i = 0; i < p_size.width; i++) { for (int j = 0; j < p_size.height; j++) { @@ -567,6 +567,8 @@ int RendererSceneRenderRD::reflection_atlas_get_size(RID p_ref_atlas) const { RID RendererSceneRenderRD::reflection_probe_instance_create(RID p_probe) { ReflectionProbeInstance rpi; rpi.probe = p_probe; + rpi.forward_id = _allocate_forward_id(FORWARD_ID_TYPE_REFLECTION_PROBE); + return reflection_probe_instance_owner.make_rid(rpi); } @@ -627,6 +629,8 @@ bool RendererSceneRenderRD::reflection_probe_instance_begin_render(RID p_instanc ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(p_instance); ERR_FAIL_COND_V(!rpi, false); + RD::get_singleton()->draw_command_begin_label("Reflection probe render"); + if (storage->reflection_probe_get_update_mode(rpi->probe) == RS::REFLECTION_PROBE_UPDATE_ALWAYS && atlas->reflection.is_valid() && atlas->size != 256) { WARN_PRINT("ReflectionProbes set to UPDATE_ALWAYS must have an atlas size of 256. Please update the atlas size in the ProjectSettings."); reflection_atlas_set_size(p_reflection_atlas, 256, atlas->count); @@ -654,12 +658,12 @@ bool RendererSceneRenderRD::reflection_probe_instance_begin_render(RID p_instanc //reflection atlas was unused, create: RD::TextureFormat tf; tf.array_layers = 6 * atlas->count; - tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + tf.format = _render_buffers_get_color_format(); tf.texture_type = RD::TEXTURE_TYPE_CUBE_ARRAY; tf.mipmaps = mipmaps; tf.width = atlas->size; tf.height = atlas->size; - tf.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + tf.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT | (_render_buffers_can_be_storage() ? RD::TEXTURE_USAGE_STORAGE_BIT : 0); atlas->reflection = RD::get_singleton()->texture_create(tf, RD::TextureView()); } @@ -673,12 +677,9 @@ bool RendererSceneRenderRD::reflection_probe_instance_begin_render(RID p_instanc } atlas->reflections.resize(atlas->count); for (int i = 0; i < atlas->count; i++) { - atlas->reflections.write[i].data.update_reflection_data(atlas->size, mipmaps, false, atlas->reflection, i * 6, storage->reflection_probe_get_update_mode(rpi->probe) == RS::REFLECTION_PROBE_UPDATE_ALWAYS, sky.roughness_layers); + atlas->reflections.write[i].data.update_reflection_data(storage, atlas->size, mipmaps, false, atlas->reflection, i * 6, storage->reflection_probe_get_update_mode(rpi->probe) == RS::REFLECTION_PROBE_UPDATE_ALWAYS, sky.roughness_layers, _render_buffers_get_color_format()); for (int j = 0; j < 6; j++) { - Vector<RID> fb; - fb.push_back(atlas->reflections.write[i].data.layers[0].mipmaps[0].views[j]); - fb.push_back(atlas->depth_buffer); - atlas->reflections.write[i].fbs[j] = RD::get_singleton()->framebuffer_create(fb); + atlas->reflections.write[i].fbs[j] = reflection_probe_create_framebuffer(atlas->reflections.write[i].data.layers[0].mipmaps[0].views[j], atlas->depth_buffer); } } @@ -709,15 +710,28 @@ bool RendererSceneRenderRD::reflection_probe_instance_begin_render(RID p_instanc } } + if (rpi->atlas_index != -1) { // should we fail if this is still -1 ? + atlas->reflections.write[rpi->atlas_index].owner = p_instance; + } + rpi->atlas = p_reflection_atlas; rpi->rendering = true; rpi->dirty = false; rpi->processing_layer = 1; rpi->processing_side = 0; + RD::get_singleton()->draw_command_end_label(); + return true; } +RID RendererSceneRenderRD::reflection_probe_create_framebuffer(RID p_color, RID p_depth) { + Vector<RID> fb; + fb.push_back(p_color); + fb.push_back(p_depth); + return RD::get_singleton()->framebuffer_create(fb); +} + bool RendererSceneRenderRD::reflection_probe_instance_postprocess_step(RID p_instance) { ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(p_instance); ERR_FAIL_COND_V(!rpi, false); @@ -846,7 +860,7 @@ void RendererSceneRenderRD::shadow_atlas_set_size(RID p_atlas, int p_size, bool shadow_atlas->shadow_owners.clear(); shadow_atlas->size = p_size; - shadow_atlas->use_16_bits = p_size; + shadow_atlas->use_16_bits = p_16_bits; } void RendererSceneRenderRD::shadow_atlas_set_quadrant_subdivision(RID p_atlas, int p_quadrant, int p_subdivision) { @@ -921,7 +935,7 @@ bool RendererSceneRenderRD::_shadow_atlas_find_shadow(ShadowAtlas *shadow_atlas, //look for an empty space int sc = shadow_atlas->quadrants[qidx].shadows.size(); - ShadowAtlas::Quadrant::Shadow *sarr = shadow_atlas->quadrants[qidx].shadows.ptrw(); + const ShadowAtlas::Quadrant::Shadow *sarr = shadow_atlas->quadrants[qidx].shadows.ptr(); int found_free_idx = -1; //found a free one int found_used_idx = -1; //found existing one, must steal it @@ -966,6 +980,78 @@ bool RendererSceneRenderRD::_shadow_atlas_find_shadow(ShadowAtlas *shadow_atlas, return false; } +bool RendererSceneRenderRD::_shadow_atlas_find_omni_shadows(ShadowAtlas *shadow_atlas, int *p_in_quadrants, int p_quadrant_count, int p_current_subdiv, uint64_t p_tick, int &r_quadrant, int &r_shadow) { + for (int i = p_quadrant_count - 1; i >= 0; i--) { + int qidx = p_in_quadrants[i]; + + if (shadow_atlas->quadrants[qidx].subdivision == (uint32_t)p_current_subdiv) { + return false; + } + + //look for an empty space + int sc = shadow_atlas->quadrants[qidx].shadows.size(); + const ShadowAtlas::Quadrant::Shadow *sarr = shadow_atlas->quadrants[qidx].shadows.ptr(); + + int found_idx = -1; + uint64_t min_pass = 0; // sum of currently selected spots, try to get the least recently used pair + + for (int j = 0; j < sc - 1; j++) { + uint64_t pass = 0; + + if (sarr[j].owner.is_valid()) { + LightInstance *sli = light_instance_owner.getornull(sarr[j].owner); + ERR_CONTINUE(!sli); + + if (sli->last_scene_pass == scene_pass) { + continue; + } + + //was just allocated, don't kill it so soon, wait a bit.. + if (p_tick - sarr[j].alloc_tick < shadow_atlas_realloc_tolerance_msec) { + continue; + } + pass += sli->last_scene_pass; + } + + if (sarr[j + 1].owner.is_valid()) { + LightInstance *sli = light_instance_owner.getornull(sarr[j + 1].owner); + ERR_CONTINUE(!sli); + + if (sli->last_scene_pass == scene_pass) { + continue; + } + + //was just allocated, don't kill it so soon, wait a bit.. + if (p_tick - sarr[j + 1].alloc_tick < shadow_atlas_realloc_tolerance_msec) { + continue; + } + pass += sli->last_scene_pass; + } + + if (found_idx == -1 || pass < min_pass) { + found_idx = j; + min_pass = pass; + + // we found two empty spots, no need to check the rest + if (pass == 0) { + break; + } + } + } + + if (found_idx == -1) { + continue; //nothing found + } + + r_quadrant = qidx; + r_shadow = found_idx; + + return true; + } + + return false; +} + bool RendererSceneRenderRD::shadow_atlas_update_light(RID p_atlas, RID p_light_intance, float p_coverage, uint64_t p_light_version) { ShadowAtlas *shadow_atlas = shadow_atlas_owner.getornull(p_atlas); ERR_FAIL_COND_V(!shadow_atlas, false); @@ -1011,94 +1097,104 @@ bool RendererSceneRenderRD::shadow_atlas_update_light(RID p_atlas, RID p_light_i uint64_t tick = OS::get_singleton()->get_ticks_msec(); - //see if it already exists + uint32_t old_key = ShadowAtlas::SHADOW_INVALID; + uint32_t old_quadrant = ShadowAtlas::SHADOW_INVALID; + uint32_t old_shadow = ShadowAtlas::SHADOW_INVALID; + int old_subdivision = -1; + + bool should_realloc = false; + bool should_redraw = false; if (shadow_atlas->shadow_owners.has(p_light_intance)) { - //it does! - uint32_t key = shadow_atlas->shadow_owners[p_light_intance]; - uint32_t q = (key >> ShadowAtlas::QUADRANT_SHIFT) & 0x3; - uint32_t s = key & ShadowAtlas::SHADOW_INDEX_MASK; + old_key = shadow_atlas->shadow_owners[p_light_intance]; + old_quadrant = (old_key >> ShadowAtlas::QUADRANT_SHIFT) & 0x3; + old_shadow = old_key & ShadowAtlas::SHADOW_INDEX_MASK; - bool should_realloc = shadow_atlas->quadrants[q].subdivision != (uint32_t)best_subdiv && (shadow_atlas->quadrants[q].shadows[s].alloc_tick - tick > shadow_atlas_realloc_tolerance_msec); - bool should_redraw = shadow_atlas->quadrants[q].shadows[s].version != p_light_version; + should_realloc = shadow_atlas->quadrants[old_quadrant].subdivision != (uint32_t)best_subdiv && (shadow_atlas->quadrants[old_quadrant].shadows[old_shadow].alloc_tick - tick > shadow_atlas_realloc_tolerance_msec); + should_redraw = shadow_atlas->quadrants[old_quadrant].shadows[old_shadow].version != p_light_version; if (!should_realloc) { - shadow_atlas->quadrants[q].shadows.write[s].version = p_light_version; + shadow_atlas->quadrants[old_quadrant].shadows.write[old_shadow].version = p_light_version; //already existing, see if it should redraw or it's just OK return should_redraw; } - int new_quadrant, new_shadow; + old_subdivision = shadow_atlas->quadrants[old_quadrant].subdivision; + } + + bool is_omni = li->light_type == RS::LIGHT_OMNI; + bool found_shadow = false; + int new_quadrant = -1; + int new_shadow = -1; - //find a better place - if (_shadow_atlas_find_shadow(shadow_atlas, valid_quadrants, valid_quadrant_count, shadow_atlas->quadrants[q].subdivision, tick, new_quadrant, new_shadow)) { - //found a better place! - ShadowAtlas::Quadrant::Shadow *sh = &shadow_atlas->quadrants[new_quadrant].shadows.write[new_shadow]; - if (sh->owner.is_valid()) { - //is taken, but is invalid, erasing it - shadow_atlas->shadow_owners.erase(sh->owner); - LightInstance *sli = light_instance_owner.getornull(sh->owner); - sli->shadow_atlases.erase(p_atlas); - } + if (is_omni) { + found_shadow = _shadow_atlas_find_omni_shadows(shadow_atlas, valid_quadrants, valid_quadrant_count, old_subdivision, tick, new_quadrant, new_shadow); + } else { + found_shadow = _shadow_atlas_find_shadow(shadow_atlas, valid_quadrants, valid_quadrant_count, old_subdivision, tick, new_quadrant, new_shadow); + } - //erase previous - shadow_atlas->quadrants[q].shadows.write[s].version = 0; - shadow_atlas->quadrants[q].shadows.write[s].owner = RID(); + if (found_shadow) { + if (old_quadrant != ShadowAtlas::SHADOW_INVALID) { + shadow_atlas->quadrants[old_quadrant].shadows.write[old_shadow].version = 0; + shadow_atlas->quadrants[old_quadrant].shadows.write[old_shadow].owner = RID(); - sh->owner = p_light_intance; - sh->alloc_tick = tick; - sh->version = p_light_version; - li->shadow_atlases.insert(p_atlas); - - //make new key - key = new_quadrant << ShadowAtlas::QUADRANT_SHIFT; - key |= new_shadow; - //update it in map - shadow_atlas->shadow_owners[p_light_intance] = key; - //make it dirty, as it should redraw anyway - return true; + if (old_key & ShadowAtlas::OMNI_LIGHT_FLAG) { + shadow_atlas->quadrants[old_quadrant].shadows.write[old_shadow + 1].version = 0; + shadow_atlas->quadrants[old_quadrant].shadows.write[old_shadow + 1].owner = RID(); + } } - //no better place for this shadow found, keep current + uint32_t new_key = new_quadrant << ShadowAtlas::QUADRANT_SHIFT; + new_key |= new_shadow; - //already existing, see if it should redraw or it's just OK + ShadowAtlas::Quadrant::Shadow *sh = &shadow_atlas->quadrants[new_quadrant].shadows.write[new_shadow]; + _shadow_atlas_invalidate_shadow(sh, p_atlas, shadow_atlas, new_quadrant, new_shadow); - shadow_atlas->quadrants[q].shadows.write[s].version = p_light_version; + sh->owner = p_light_intance; + sh->alloc_tick = tick; + sh->version = p_light_version; - return should_redraw; - } + if (is_omni) { + new_key |= ShadowAtlas::OMNI_LIGHT_FLAG; - int new_quadrant, new_shadow; + int new_omni_shadow = new_shadow + 1; + ShadowAtlas::Quadrant::Shadow *extra_sh = &shadow_atlas->quadrants[new_quadrant].shadows.write[new_omni_shadow]; + _shadow_atlas_invalidate_shadow(extra_sh, p_atlas, shadow_atlas, new_quadrant, new_omni_shadow); - //find a better place - if (_shadow_atlas_find_shadow(shadow_atlas, valid_quadrants, valid_quadrant_count, -1, tick, new_quadrant, new_shadow)) { - //found a better place! - ShadowAtlas::Quadrant::Shadow *sh = &shadow_atlas->quadrants[new_quadrant].shadows.write[new_shadow]; - if (sh->owner.is_valid()) { - //is taken, but is invalid, erasing it - shadow_atlas->shadow_owners.erase(sh->owner); - LightInstance *sli = light_instance_owner.getornull(sh->owner); - sli->shadow_atlases.erase(p_atlas); + extra_sh->owner = p_light_intance; + extra_sh->alloc_tick = tick; + extra_sh->version = p_light_version; } - sh->owner = p_light_intance; - sh->alloc_tick = tick; - sh->version = p_light_version; li->shadow_atlases.insert(p_atlas); - //make new key - uint32_t key = new_quadrant << ShadowAtlas::QUADRANT_SHIFT; - key |= new_shadow; //update it in map - shadow_atlas->shadow_owners[p_light_intance] = key; + shadow_atlas->shadow_owners[p_light_intance] = new_key; //make it dirty, as it should redraw anyway - return true; } - //no place to allocate this light, apologies + return should_redraw; +} + +void RendererSceneRenderRD::_shadow_atlas_invalidate_shadow(RendererSceneRenderRD::ShadowAtlas::Quadrant::Shadow *p_shadow, RID p_atlas, RendererSceneRenderRD::ShadowAtlas *p_shadow_atlas, uint32_t p_quadrant, uint32_t p_shadow_idx) { + if (p_shadow->owner.is_valid()) { + LightInstance *sli = light_instance_owner.getornull(p_shadow->owner); + uint32_t old_key = p_shadow_atlas->shadow_owners[p_shadow->owner]; - return false; + if (old_key & ShadowAtlas::OMNI_LIGHT_FLAG) { + uint32_t s = old_key & ShadowAtlas::SHADOW_INDEX_MASK; + uint32_t omni_shadow_idx = p_shadow_idx + (s == (uint32_t)p_shadow_idx ? 1 : -1); + RendererSceneRenderRD::ShadowAtlas::Quadrant::Shadow *omni_shadow = &p_shadow_atlas->quadrants[p_quadrant].shadows.write[omni_shadow_idx]; + omni_shadow->version = 0; + omni_shadow->owner = RID(); + } + + p_shadow->version = 0; + p_shadow->owner = RID(); + sli->shadow_atlases.erase(p_atlas); + p_shadow_atlas->shadow_owners.erase(p_shadow->owner); + } } void RendererSceneRenderRD::_update_directional_shadow_atlas() { @@ -1123,6 +1219,7 @@ void RendererSceneRenderRD::directional_shadow_atlas_set_size(int p_size, bool p } directional_shadow.size = p_size; + directional_shadow.use_16_bits = p_16_bits; if (directional_shadow.depth.is_valid()) { RD::get_singleton()->free(directional_shadow.depth); @@ -1229,6 +1326,9 @@ RID RendererSceneRenderRD::light_instance_create(RID p_light) { light_instance->self = li; light_instance->light = p_light; light_instance->light_type = storage->light_get_type(p_light); + if (light_instance->light_type != RS::LIGHT_DIRECTIONAL) { + light_instance->forward_id = _allocate_forward_id(light_instance->light_type == RS::LIGHT_OMNI ? FORWARD_ID_TYPE_OMNI_LIGHT : FORWARD_ID_TYPE_SPOT_LIGHT); + } return li; } @@ -1302,6 +1402,7 @@ RendererSceneRenderRD::ShadowCubemap *RendererSceneRenderRD::_get_shadow_cubemap RID RendererSceneRenderRD::decal_instance_create(RID p_decal) { DecalInstance di; di.decal = p_decal; + di.forward_id = _allocate_forward_id(FORWARD_ID_TYPE_DECAL); return decal_instance_owner.make_rid(di); } @@ -1373,12 +1474,20 @@ void RendererSceneRenderRD::_allocate_blur_textures(RenderBuffers *rb) { uint32_t mipmaps_required = Image::get_image_required_mipmaps(rb->width, rb->height, Image::FORMAT_RGBAH); + // TODO make sure texture_create_shared_from_slice works for multiview + RD::TextureFormat tf; - tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + tf.format = _render_buffers_get_color_format(); // RD::DATA_FORMAT_R16G16B16A16_SFLOAT; tf.width = rb->width; tf.height = rb->height; - tf.texture_type = RD::TEXTURE_TYPE_2D; - tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT; + tf.texture_type = rb->view_count > 1 ? RD::TEXTURE_TYPE_2D_ARRAY : RD::TEXTURE_TYPE_2D; + tf.array_layers = rb->view_count; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT; + if (_render_buffers_can_be_storage()) { + tf.usage_bits += RD::TEXTURE_USAGE_STORAGE_BIT; + } else { + tf.usage_bits += RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; + } tf.mipmaps = mipmaps_required; rb->blur[0].texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); @@ -1398,17 +1507,125 @@ void RendererSceneRenderRD::_allocate_blur_textures(RenderBuffers *rb) { mm.width = base_width; mm.height = base_height; + if (!_render_buffers_can_be_storage()) { + Vector<RID> fb; + fb.push_back(mm.texture); + + mm.fb = RD::get_singleton()->framebuffer_create(fb); + } + + if (!_render_buffers_can_be_storage()) { + // and half texture, this is an intermediate result so just allocate a texture, is this good enough? + tf.width = MAX(1, base_width >> 1); + tf.height = base_height; + tf.mipmaps = 1; // 1 or 0? + + mm.half_texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); + + Vector<RID> half_fb; + half_fb.push_back(mm.half_texture); + mm.half_fb = RD::get_singleton()->framebuffer_create(half_fb); + } + rb->blur[0].mipmaps.push_back(mm); if (i > 0) { mm.texture = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), rb->blur[1].texture, 0, i - 1); + if (!_render_buffers_can_be_storage()) { + Vector<RID> fb; + fb.push_back(mm.texture); + + mm.fb = RD::get_singleton()->framebuffer_create(fb); + + // We can re-use the half texture here as it is an intermediate result + } + rb->blur[1].mipmaps.push_back(mm); } base_width = MAX(1, base_width >> 1); base_height = MAX(1, base_height >> 1); } + + if (!_render_buffers_can_be_storage()) { + // create 4 weight textures, 2 full size, 2 half size + + tf.format = RD::DATA_FORMAT_R16_SFLOAT; // We could probably use DATA_FORMAT_R8_SNORM if we don't pre-multiply by blur_size but that depends on whether we can remove DEPTH_GAP + tf.width = rb->width; + tf.height = rb->height; + tf.texture_type = rb->view_count > 1 ? RD::TEXTURE_TYPE_2D_ARRAY : RD::TEXTURE_TYPE_2D; + tf.array_layers = rb->view_count; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_CAN_COPY_TO_BIT | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; + tf.mipmaps = 1; + for (uint32_t i = 0; i < 4; i++) { + // associated blur texture + RID texture; + if (i == 0) { + texture = rb->texture; + } else if (i == 1) { + texture = rb->blur[0].mipmaps[0].texture; + } else if (i == 2) { + texture = rb->blur[1].mipmaps[0].texture; + } else if (i == 3) { + texture = rb->blur[0].mipmaps[1].texture; + } + + // create weight texture + rb->weight_buffers[i].weight = RD::get_singleton()->texture_create(tf, RD::TextureView()); + + // create frame buffer + Vector<RID> fb; + fb.push_back(texture); + fb.push_back(rb->weight_buffers[i].weight); + rb->weight_buffers[i].fb = RD::get_singleton()->framebuffer_create(fb); + + if (i == 1) { + // next 2 are half size + tf.width = MAX(1, tf.width >> 1); + tf.height = MAX(1, tf.height >> 1); + } + } + + { + // and finally an FB for just our base weights + Vector<RID> fb; + fb.push_back(rb->weight_buffers[0].weight); + rb->base_weight_fb = RD::get_singleton()->framebuffer_create(fb); + } + } +} + +void RendererSceneRenderRD::_allocate_depth_backbuffer_textures(RenderBuffers *rb) { + ERR_FAIL_COND(!rb->depth_back_texture.is_null()); + + { + RD::TextureFormat tf; + if (rb->view_count > 1) { + tf.texture_type = RD::TEXTURE_TYPE_2D_ARRAY; + } + // We're not using this as a depth stencil, just copying our data into this. May need to look into using a different format on mobile, maybe R16? + tf.format = RD::DATA_FORMAT_R32_SFLOAT; + + tf.width = rb->width; + tf.height = rb->height; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT; + tf.array_layers = rb->view_count; // create a layer for every view + + tf.usage_bits |= RD::TEXTURE_USAGE_CAN_COPY_TO_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + tf.usage_bits |= RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; // set this as color attachment because we're copying data into it, it's not actually used as a depth buffer + + rb->depth_back_texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); + } + + if (!_render_buffers_can_be_storage()) { + // create framebuffer so we can write into this... + + Vector<RID> fb; + fb.push_back(rb->depth_back_texture); + + rb->depth_back_fb = RD::get_singleton()->framebuffer_create(fb, RD::INVALID_ID, rb->view_count); + } } void RendererSceneRenderRD::_allocate_luminance_textures(RenderBuffers *rb) { @@ -1425,26 +1642,48 @@ void RendererSceneRenderRD::_allocate_luminance_textures(RenderBuffers *rb) { tf.format = RD::DATA_FORMAT_R32_SFLOAT; tf.width = w; tf.height = h; - tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT; bool final = w == 1 && h == 1; - if (final) { - tf.usage_bits |= RD::TEXTURE_USAGE_SAMPLING_BIT; + if (_render_buffers_can_be_storage()) { + tf.usage_bits = RD::TEXTURE_USAGE_STORAGE_BIT; + if (final) { + tf.usage_bits |= RD::TEXTURE_USAGE_SAMPLING_BIT; + } + } else { + tf.usage_bits = RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT | RD::TEXTURE_USAGE_SAMPLING_BIT; } RID texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); rb->luminance.reduce.push_back(texture); + if (!_render_buffers_can_be_storage()) { + Vector<RID> fb; + fb.push_back(texture); + + rb->luminance.fb.push_back(RD::get_singleton()->framebuffer_create(fb)); + } if (final) { rb->luminance.current = RD::get_singleton()->texture_create(tf, RD::TextureView()); + + if (!_render_buffers_can_be_storage()) { + Vector<RID> fb; + fb.push_back(rb->luminance.current); + + rb->luminance.current_fb = RD::get_singleton()->framebuffer_create(fb); + } break; } } } void RendererSceneRenderRD::_free_render_buffer_data(RenderBuffers *rb) { + if (rb->texture_fb.is_valid()) { + RD::get_singleton()->free(rb->texture_fb); + rb->texture_fb = RID(); + } + if (rb->texture.is_valid()) { RD::get_singleton()->free(rb->texture); rb->texture = RID(); @@ -1455,20 +1694,54 @@ void RendererSceneRenderRD::_free_render_buffer_data(RenderBuffers *rb) { rb->depth_texture = RID(); } + if (rb->depth_back_fb.is_valid()) { + RD::get_singleton()->free(rb->depth_back_fb); + rb->depth_back_fb = RID(); + } + + if (rb->depth_back_texture.is_valid()) { + RD::get_singleton()->free(rb->depth_back_texture); + rb->depth_back_texture = RID(); + } + for (int i = 0; i < 2; i++) { + for (int m = 0; m < rb->blur[i].mipmaps.size(); m++) { + // do we free the texture slice here? or is it enough to free the main texture? + + // do free the mobile extra stuff + if (rb->blur[i].mipmaps[m].fb.is_valid()) { + RD::get_singleton()->free(rb->blur[i].mipmaps[m].fb); + } + if (rb->blur[i].mipmaps[m].half_fb.is_valid()) { + RD::get_singleton()->free(rb->blur[i].mipmaps[m].half_fb); + } + if (rb->blur[i].mipmaps[m].half_texture.is_valid()) { + RD::get_singleton()->free(rb->blur[i].mipmaps[m].half_texture); + } + } + rb->blur[i].mipmaps.clear(); + if (rb->blur[i].texture.is_valid()) { RD::get_singleton()->free(rb->blur[i].texture); rb->blur[i].texture = RID(); - rb->blur[i].mipmaps.clear(); } } + for (int i = 0; i < rb->luminance.fb.size(); i++) { + RD::get_singleton()->free(rb->luminance.fb[i]); + } + rb->luminance.fb.clear(); + for (int i = 0; i < rb->luminance.reduce.size(); i++) { RD::get_singleton()->free(rb->luminance.reduce[i]); } - rb->luminance.reduce.clear(); + if (rb->luminance.current_fb.is_valid()) { + RD::get_singleton()->free(rb->luminance.current_fb); + rb->luminance.current_fb = RID(); + } + if (rb->luminance.current.is_valid()) { RD::get_singleton()->free(rb->luminance.current); rb->luminance.current = RID(); @@ -1728,7 +2001,59 @@ void RendererSceneRenderRD::_process_ssao(RID p_render_buffers, RID p_environmen settings.half_screen_size = Size2i(buffer_width, buffer_height); settings.quarter_screen_size = Size2i(half_width, half_height); - storage->get_effects()->generate_ssao(rb->depth_texture, p_normal_buffer, rb->ssao.depth, rb->ssao.depth_slices, rb->ssao.ao_deinterleaved, rb->ssao.ao_deinterleaved_slices, rb->ssao.ao_pong, rb->ssao.ao_pong_slices, rb->ssao.ao_final, rb->ssao.importance_map[0], rb->ssao.importance_map[1], p_projection, settings, uniform_sets_are_invalid); + storage->get_effects()->generate_ssao(rb->depth_texture, p_normal_buffer, rb->ssao.depth, rb->ssao.depth_slices, rb->ssao.ao_deinterleaved, rb->ssao.ao_deinterleaved_slices, rb->ssao.ao_pong, rb->ssao.ao_pong_slices, rb->ssao.ao_final, rb->ssao.importance_map[0], rb->ssao.importance_map[1], p_projection, settings, uniform_sets_are_invalid, rb->ssao.downsample_uniform_set, rb->ssao.gather_uniform_set, rb->ssao.importance_map_uniform_set); +} + +void RendererSceneRenderRD::_render_buffers_copy_screen_texture(const RenderDataRD *p_render_data) { + RenderBuffers *rb = render_buffers_owner.getornull(p_render_data->render_buffers); + ERR_FAIL_COND(!rb); + + RD::get_singleton()->draw_command_begin_label("Copy screen texture"); + + if (rb->blur[0].texture.is_null()) { + _allocate_blur_textures(rb); + } + + // @TODO IMPLEMENT MULTIVIEW, all effects need to support stereo buffers or effects are only applied to the left eye + + bool can_use_storage = _render_buffers_can_be_storage(); + + if (can_use_storage) { + storage->get_effects()->copy_to_rect(rb->texture, rb->blur[0].mipmaps[0].texture, Rect2i(0, 0, rb->width, rb->height)); + for (int i = 1; i < rb->blur[0].mipmaps.size(); i++) { + storage->get_effects()->make_mipmap(rb->blur[0].mipmaps[i - 1].texture, rb->blur[0].mipmaps[i].texture, Size2i(rb->blur[0].mipmaps[i].width, rb->blur[0].mipmaps[i].height)); + } + } else { + storage->get_effects()->copy_to_fb_rect(rb->texture, rb->blur[0].mipmaps[0].fb, Rect2i(0, 0, rb->width, rb->height)); + for (int i = 1; i < rb->blur[0].mipmaps.size(); i++) { + storage->get_effects()->make_mipmap_raster(rb->blur[0].mipmaps[i - 1].texture, rb->blur[0].mipmaps[i].fb, Size2i(rb->blur[0].mipmaps[i].width, rb->blur[0].mipmaps[i].height)); + } + } + + RD::get_singleton()->draw_command_end_label(); +} + +void RendererSceneRenderRD::_render_buffers_copy_depth_texture(const RenderDataRD *p_render_data) { + RenderBuffers *rb = render_buffers_owner.getornull(p_render_data->render_buffers); + ERR_FAIL_COND(!rb); + + RD::get_singleton()->draw_command_begin_label("Copy depth texture"); + + if (rb->depth_back_texture.is_null()) { + _allocate_depth_backbuffer_textures(rb); + } + + // @TODO IMPLEMENT MULTIVIEW, all effects need to support stereo buffers or effects are only applied to the left eye + + bool can_use_storage = _render_buffers_can_be_storage(); + + if (can_use_storage) { + storage->get_effects()->copy_to_rect(rb->depth_texture, rb->depth_back_texture, Rect2i(0, 0, rb->width, rb->height)); + } else { + storage->get_effects()->copy_to_fb_rect(rb->depth_texture, rb->depth_back_fb, Rect2i(0, 0, rb->width, rb->height)); + } + + RD::get_singleton()->draw_command_end_label(); } void RendererSceneRenderRD::_render_buffers_post_process_and_tonemap(const RenderDataRD *p_render_data) { @@ -1740,17 +2065,50 @@ void RendererSceneRenderRD::_render_buffers_post_process_and_tonemap(const Rende CameraEffects *camfx = camera_effects_owner.getornull(p_render_data->camera_effects); bool can_use_effects = rb->width >= 8 && rb->height >= 8; + bool can_use_storage = _render_buffers_can_be_storage(); + + // @TODO IMPLEMENT MULTIVIEW, all effects need to support stereo buffers or effects are only applied to the left eye if (can_use_effects && camfx && (camfx->dof_blur_near_enabled || camfx->dof_blur_far_enabled) && camfx->dof_blur_amount > 0.0) { + RD::get_singleton()->draw_command_begin_label("DOF"); if (rb->blur[0].texture.is_null()) { _allocate_blur_textures(rb); } + EffectsRD::BokehBuffers buffers; + + // textures we use + buffers.base_texture_size = Size2i(rb->width, rb->height); + buffers.base_texture = rb->texture; + buffers.depth_texture = rb->depth_texture; + buffers.secondary_texture = rb->blur[0].mipmaps[0].texture; + buffers.half_texture[0] = rb->blur[1].mipmaps[0].texture; + buffers.half_texture[1] = rb->blur[0].mipmaps[1].texture; + float bokeh_size = camfx->dof_blur_amount * 64.0; - storage->get_effects()->bokeh_dof(rb->texture, rb->depth_texture, Size2i(rb->width, rb->height), rb->blur[0].mipmaps[0].texture, rb->blur[1].mipmaps[0].texture, rb->blur[0].mipmaps[1].texture, camfx->dof_blur_far_enabled, camfx->dof_blur_far_distance, camfx->dof_blur_far_transition, camfx->dof_blur_near_enabled, camfx->dof_blur_near_distance, camfx->dof_blur_near_transition, bokeh_size, dof_blur_bokeh_shape, dof_blur_quality, dof_blur_use_jitter, p_render_data->z_near, p_render_data->z_far, p_render_data->cam_ortogonal); + if (can_use_storage) { + storage->get_effects()->bokeh_dof(buffers, camfx->dof_blur_far_enabled, camfx->dof_blur_far_distance, camfx->dof_blur_far_transition, camfx->dof_blur_near_enabled, camfx->dof_blur_near_distance, camfx->dof_blur_near_transition, bokeh_size, dof_blur_bokeh_shape, dof_blur_quality, dof_blur_use_jitter, p_render_data->z_near, p_render_data->z_far, p_render_data->cam_ortogonal); + } else { + // set framebuffers + buffers.base_fb = rb->texture_fb; + buffers.secondary_fb = rb->weight_buffers[1].fb; + buffers.half_fb[0] = rb->weight_buffers[2].fb; + buffers.half_fb[1] = rb->weight_buffers[3].fb; + buffers.weight_texture[0] = rb->weight_buffers[0].weight; + buffers.weight_texture[1] = rb->weight_buffers[1].weight; + buffers.weight_texture[2] = rb->weight_buffers[2].weight; + buffers.weight_texture[3] = rb->weight_buffers[3].weight; + + // set weight buffers + buffers.base_weight_fb = rb->base_weight_fb; + + storage->get_effects()->bokeh_dof_raster(buffers, camfx->dof_blur_far_enabled, camfx->dof_blur_far_distance, camfx->dof_blur_far_transition, camfx->dof_blur_near_enabled, camfx->dof_blur_near_distance, camfx->dof_blur_near_transition, bokeh_size, dof_blur_bokeh_shape, dof_blur_quality, p_render_data->z_near, p_render_data->z_far, p_render_data->cam_ortogonal); + } + RD::get_singleton()->draw_command_end_label(); } if (can_use_effects && env && env->auto_exposure) { + RD::get_singleton()->draw_command_begin_label("Auto exposure"); if (rb->luminance.current.is_null()) { _allocate_luminance_textures(rb); } @@ -1759,16 +2117,26 @@ void RendererSceneRenderRD::_render_buffers_post_process_and_tonemap(const Rende rb->auto_exposure_version = env->auto_exposure_version; double step = env->auto_exp_speed * time_step; - storage->get_effects()->luminance_reduction(rb->texture, Size2i(rb->width, rb->height), rb->luminance.reduce, rb->luminance.current, env->min_luminance, env->max_luminance, step, set_immediate); - + if (can_use_storage) { + storage->get_effects()->luminance_reduction(rb->texture, Size2i(rb->width, rb->height), rb->luminance.reduce, rb->luminance.current, env->min_luminance, env->max_luminance, step, set_immediate); + } else { + storage->get_effects()->luminance_reduction_raster(rb->texture, Size2i(rb->width, rb->height), rb->luminance.reduce, rb->luminance.fb, rb->luminance.current, env->min_luminance, env->max_luminance, step, set_immediate); + } //swap final reduce with prev luminance SWAP(rb->luminance.current, rb->luminance.reduce.write[rb->luminance.reduce.size() - 1]); + if (!can_use_storage) { + SWAP(rb->luminance.current_fb, rb->luminance.fb.write[rb->luminance.fb.size() - 1]); + } + RenderingServerDefault::redraw_request(); //redraw all the time if auto exposure rendering is on + RD::get_singleton()->draw_command_end_label(); } int max_glow_level = -1; if (can_use_effects && env && env->glow_enabled) { + RD::get_singleton()->draw_command_begin_label("Gaussian Glow"); + /* see that blur textures are allocated */ if (rb->blur[1].texture.is_null()) { @@ -1794,14 +2162,26 @@ void RendererSceneRenderRD::_render_buffers_post_process_and_tonemap(const Rende if (env->auto_exposure && rb->luminance.current.is_valid()) { luminance_texture = rb->luminance.current; } - storage->get_effects()->gaussian_glow(rb->texture, rb->blur[1].mipmaps[i].texture, Size2i(vp_w, vp_h), env->glow_strength, glow_high_quality, true, env->glow_hdr_luminance_cap, env->exposure, env->glow_bloom, env->glow_hdr_bleed_threshold, env->glow_hdr_bleed_scale, luminance_texture, env->auto_exp_scale); + if (can_use_storage) { + storage->get_effects()->gaussian_glow(rb->texture, rb->blur[1].mipmaps[i].texture, Size2i(vp_w, vp_h), env->glow_strength, glow_high_quality, true, env->glow_hdr_luminance_cap, env->exposure, env->glow_bloom, env->glow_hdr_bleed_threshold, env->glow_hdr_bleed_scale, luminance_texture, env->auto_exp_scale); + } else { + storage->get_effects()->gaussian_glow_raster(rb->texture, rb->blur[1].mipmaps[i].half_fb, rb->blur[1].mipmaps[i].half_texture, rb->blur[1].mipmaps[i].fb, Size2i(vp_w, vp_h), env->glow_strength, glow_high_quality, true, env->glow_hdr_luminance_cap, env->exposure, env->glow_bloom, env->glow_hdr_bleed_threshold, env->glow_hdr_bleed_scale, luminance_texture, env->auto_exp_scale); + } } else { - storage->get_effects()->gaussian_glow(rb->blur[1].mipmaps[i - 1].texture, rb->blur[1].mipmaps[i].texture, Size2i(vp_w, vp_h), env->glow_strength, glow_high_quality); + if (can_use_storage) { + storage->get_effects()->gaussian_glow(rb->blur[1].mipmaps[i - 1].texture, rb->blur[1].mipmaps[i].texture, Size2i(vp_w, vp_h), env->glow_strength, glow_high_quality); + } else { + storage->get_effects()->gaussian_glow_raster(rb->blur[1].mipmaps[i - 1].texture, rb->blur[1].mipmaps[i].half_fb, rb->blur[1].mipmaps[i].half_texture, rb->blur[1].mipmaps[i].fb, Vector2(1.0 / vp_w, 1.0 / vp_h), env->glow_strength, glow_high_quality); + } } } + + RD::get_singleton()->draw_command_end_label(); } { + RD::get_singleton()->draw_command_begin_label("Tonemap"); + //tonemap EffectsRD::TonemapSettings tonemap; @@ -1857,11 +2237,83 @@ void RendererSceneRenderRD::_render_buffers_post_process_and_tonemap(const Rende } } + tonemap.luminance_multiplier = _render_buffers_get_luminance_multiplier(); tonemap.view_count = p_render_data->view_count; storage->get_effects()->tonemapper(rb->texture, storage->render_target_get_rd_framebuffer(rb->render_target), tonemap); + + RD::get_singleton()->draw_command_end_label(); + } + + storage->render_target_disable_clear_request(rb->render_target); +} + +void RendererSceneRenderRD::_post_process_subpass(RID p_source_texture, RID p_framebuffer, const RenderDataRD *p_render_data) { + RD::get_singleton()->draw_command_begin_label("Post Process Subpass"); + + RenderBuffers *rb = render_buffers_owner.getornull(p_render_data->render_buffers); + ERR_FAIL_COND(!rb); + + RendererSceneEnvironmentRD *env = environment_owner.getornull(p_render_data->environment); + + bool can_use_effects = rb->width >= 8 && rb->height >= 8; + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_switch_to_next_pass(); + + EffectsRD::TonemapSettings tonemap; + + if (env) { + tonemap.tonemap_mode = env->tone_mapper; + tonemap.exposure = env->exposure; + tonemap.white = env->white; } + // We don't support glow or auto exposure here, if they are needed, don't use subpasses! + // The problem is that we need to use the result so far and process them before we can + // apply this to our results. + if (can_use_effects && env && env->glow_enabled) { + ERR_FAIL_MSG("Glow is not supported when using subpasses."); + } + if (can_use_effects && env && env->auto_exposure) { + ERR_FAIL_MSG("Glow is not supported when using subpasses."); + } + + tonemap.use_glow = false; + tonemap.glow_texture = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_BLACK); + tonemap.use_auto_exposure = false; + tonemap.exposure_texture = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_WHITE); + + tonemap.use_color_correction = false; + tonemap.use_1d_color_correction = false; + tonemap.color_correction_texture = storage->texture_rd_get_default(RendererStorageRD::DEFAULT_RD_TEXTURE_3D_WHITE); + + if (can_use_effects && env) { + tonemap.use_bcs = env->adjustments_enabled; + tonemap.brightness = env->adjustments_brightness; + tonemap.contrast = env->adjustments_contrast; + tonemap.saturation = env->adjustments_saturation; + if (env->adjustments_enabled && env->color_correction.is_valid()) { + tonemap.use_color_correction = true; + tonemap.use_1d_color_correction = env->use_1d_color_correction; + tonemap.color_correction_texture = storage->texture_get_rd_texture(env->color_correction); + } + } + + tonemap.use_debanding = rb->use_debanding; + tonemap.texture_size = Vector2i(rb->width, rb->height); + + tonemap.luminance_multiplier = _render_buffers_get_luminance_multiplier(); + tonemap.view_count = p_render_data->view_count; + + storage->get_effects()->tonemapper(draw_list, p_source_texture, RD::get_singleton()->framebuffer_get_format(p_framebuffer), tonemap); + + RD::get_singleton()->draw_command_end_label(); +} + +void RendererSceneRenderRD::_disable_clear_request(const RenderDataRD *p_render_data) { + RenderBuffers *rb = render_buffers_owner.getornull(p_render_data->render_buffers); + ERR_FAIL_COND(!rb); + storage->render_target_disable_clear_request(rb->render_target); } @@ -1954,6 +2406,15 @@ RID RendererSceneRenderRD::render_buffers_get_back_buffer_texture(RID p_render_b return rb->blur[0].texture; } +RID RendererSceneRenderRD::render_buffers_get_back_depth_texture(RID p_render_buffers) { + RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); + ERR_FAIL_COND_V(!rb, RID()); + if (!rb->depth_back_texture.is_valid()) { + return RID(); //not valid at the moment + } + return rb->depth_back_texture; +} + RID RendererSceneRenderRD::render_buffers_get_ao_texture(RID p_render_buffers) { RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); ERR_FAIL_COND_V(!rb, RID()); @@ -2114,10 +2575,24 @@ float RendererSceneRenderRD::render_buffers_get_volumetric_fog_detail_spread(RID return rb->volumetric_fog->spread; } +float RendererSceneRenderRD::_render_buffers_get_luminance_multiplier() { + return 1.0; +} + +RD::DataFormat RendererSceneRenderRD::_render_buffers_get_color_format() { + return RD::DATA_FORMAT_R16G16B16A16_SFLOAT; +} + +bool RendererSceneRenderRD::_render_buffers_can_be_storage() { + return true; +} + void RendererSceneRenderRD::render_buffers_configure(RID p_render_buffers, RID p_render_target, int p_width, int p_height, RS::ViewportMSAA p_msaa, RenderingServer::ViewportScreenSpaceAA p_screen_space_aa, bool p_use_debanding, uint32_t p_view_count) { - ERR_FAIL_COND_MSG(p_view_count == 0, "Must have atleast 1 view"); + ERR_FAIL_COND_MSG(p_view_count == 0, "Must have at least 1 view"); RenderBuffers *rb = render_buffers_owner.getornull(p_render_buffers); + + // Should we add an overrule per viewport? rb->width = p_width; rb->height = p_height; rb->render_target = p_render_target; @@ -2140,16 +2615,15 @@ void RendererSceneRenderRD::render_buffers_configure(RID p_render_buffers, RID p if (rb->view_count > 1) { tf.texture_type = RD::TEXTURE_TYPE_2D_ARRAY; } - tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + tf.format = _render_buffers_get_color_format(); tf.width = rb->width; tf.height = rb->height; tf.array_layers = rb->view_count; // create a layer for every view - tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; + tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | (_render_buffers_can_be_storage() ? RD::TEXTURE_USAGE_STORAGE_BIT : 0) | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; if (rb->msaa != RS::VIEWPORT_MSAA_DISABLED) { - tf.usage_bits |= RD::TEXTURE_USAGE_CAN_COPY_TO_BIT | RD::TEXTURE_USAGE_STORAGE_BIT; - } else { - tf.usage_bits |= RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; + tf.usage_bits |= RD::TEXTURE_USAGE_CAN_COPY_TO_BIT; } + tf.usage_bits |= RD::TEXTURE_USAGE_INPUT_ATTACHMENT_BIT; // only needed when using subpasses in the mobile renderer rb->texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); } @@ -2165,8 +2639,8 @@ void RendererSceneRenderRD::render_buffers_configure(RID p_render_buffers, RID p tf.format = RD::DATA_FORMAT_R32_SFLOAT; } - tf.width = p_width; - tf.height = p_height; + tf.width = rb->width; + tf.height = rb->height; tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT; tf.array_layers = rb->view_count; // create a layer for every view @@ -2179,10 +2653,19 @@ void RendererSceneRenderRD::render_buffers_configure(RID p_render_buffers, RID p rb->depth_texture = RD::get_singleton()->texture_create(tf, RD::TextureView()); } - rb->data->configure(rb->texture, rb->depth_texture, p_width, p_height, p_msaa, p_view_count); + if (!_render_buffers_can_be_storage()) { + // ONLY USED ON MOBILE RENDERER, ONLY USED FOR POST EFFECTS! + Vector<RID> fb; + fb.push_back(rb->texture); + + rb->texture_fb = RD::get_singleton()->framebuffer_create(fb, RenderingDevice::INVALID_ID, rb->view_count); + } + + RID target_texture = storage->render_target_get_rd_texture(rb->render_target); + rb->data->configure(rb->texture, rb->depth_texture, target_texture, rb->width, rb->height, p_msaa, p_view_count); if (is_clustered_enabled()) { - rb->cluster_builder->setup(Size2i(p_width, p_height), max_cluster_elements, rb->depth_texture, storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED), rb->texture); + rb->cluster_builder->setup(Size2i(rb->width, rb->height), max_cluster_elements, rb->depth_texture, storage->sampler_rd_get_default(RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST, RS::CANVAS_ITEM_TEXTURE_REPEAT_DISABLED), rb->texture); } } @@ -2241,6 +2724,8 @@ void RendererSceneRenderRD::shadows_quality_set(RS::ShadowQuality p_quality) { get_vogel_disk(penumbra_shadow_kernel, penumbra_shadow_samples); get_vogel_disk(soft_shadow_kernel, soft_shadow_samples); } + + _update_shader_quality_settings(); } void RendererSceneRenderRD::directional_shadow_quality_set(RS::ShadowQuality p_quality) { @@ -2281,6 +2766,23 @@ void RendererSceneRenderRD::directional_shadow_quality_set(RS::ShadowQuality p_q get_vogel_disk(directional_penumbra_shadow_kernel, directional_penumbra_shadow_samples); get_vogel_disk(directional_soft_shadow_kernel, directional_soft_shadow_samples); } + + _update_shader_quality_settings(); +} + +void RendererSceneRenderRD::decals_set_filter(RenderingServer::DecalFilter p_filter) { + if (decals_filter == p_filter) { + return; + } + decals_filter = p_filter; + _update_shader_quality_settings(); +} +void RendererSceneRenderRD::light_projectors_set_filter(RenderingServer::LightProjectorFilter p_filter) { + if (light_projectors_filter == p_filter) { + return; + } + light_projectors_filter = p_filter; + _update_shader_quality_settings(); } int RendererSceneRenderRD::get_roughness_layers() const { @@ -2320,10 +2822,13 @@ void RendererSceneRenderRD::_setup_reflections(const PagedArray<RID> &p_reflecti sort_array.sort(cluster.reflection_sort, cluster.reflection_count); } + bool using_forward_ids = _uses_forward_ids(); for (uint32_t i = 0; i < cluster.reflection_count; i++) { ReflectionProbeInstance *rpi = cluster.reflection_sort[i].instance; - rpi->render_index = i; + if (using_forward_ids) { + _map_forward_id(FORWARD_ID_TYPE_REFLECTION_PROBE, rpi->forward_id, i); + } RID base_probe = rpi->probe; @@ -2373,7 +2878,7 @@ void RendererSceneRenderRD::_setup_reflections(const PagedArray<RID> &p_reflecti } } -void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const Transform3D &p_camera_transform, RID p_shadow_atlas, bool p_using_shadows, uint32_t &r_directional_light_count, uint32_t &r_positional_light_count) { +void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const Transform3D &p_camera_transform, RID p_shadow_atlas, bool p_using_shadows, uint32_t &r_directional_light_count, uint32_t &r_positional_light_count, bool &r_directional_light_soft_shadows) { Transform3D inverse_transform = p_camera_transform.affine_inverse(); r_directional_light_count = 0; @@ -2385,6 +2890,8 @@ void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const cluster.omni_light_count = 0; cluster.spot_light_count = 0; + r_directional_light_soft_shadows = false; + for (int i = 0; i < (int)p_lights.size(); i++) { LightInstance *li = light_instance_owner.getornull(p_lights[i]); if (!li) { @@ -2423,6 +2930,9 @@ void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const // technically this will keep expanding until reaching the sun, but all we care // is expand until we reach the radius of the near plane (there can't be more occluders than that) angular_diameter = Math::tan(Math::deg2rad(angular_diameter)); + if (storage->light_has_shadow(base)) { + r_directional_light_soft_shadows = true; + } } else { angular_diameter = 0.0; } @@ -2531,7 +3041,7 @@ void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const CameraMatrix shadow_mtx = rectm * bias * matrix * modelview; light_data.shadow_split_offsets[j] = split; float bias_scale = li->shadow_transform[j].bias_scale; - light_data.shadow_bias[j] = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_BIAS) * bias_scale; + light_data.shadow_bias[j] = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_BIAS) / 100.0 * bias_scale; light_data.shadow_normal_bias[j] = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_NORMAL_BIAS) * li->shadow_transform[j].shadow_texel_size; light_data.shadow_transmittance_bias[j] = storage->light_get_transmittance_bias(base) * bias_scale; light_data.shadow_z_range[j] = li->shadow_transform[j].farplane; @@ -2615,6 +3125,8 @@ void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const shadow_atlas = shadow_atlas_owner.getornull(p_shadow_atlas); } + bool using_forward_ids = _uses_forward_ids(); + for (uint32_t i = 0; i < (cluster.omni_light_count + cluster.spot_light_count); i++) { uint32_t index = (i < cluster.omni_light_count) ? i : i - (cluster.omni_light_count); Cluster::LightData &light_data = (i < cluster.omni_light_count) ? cluster.omni_lights[index] : cluster.spot_lights[index]; @@ -2622,6 +3134,10 @@ void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const LightInstance *li = (i < cluster.omni_light_count) ? cluster.omni_light_sort[index].instance : cluster.spot_light_sort[index].instance; RID base = li->light; + if (using_forward_ids) { + _map_forward_id(type == RS::LIGHT_OMNI ? FORWARD_ID_TYPE_OMNI_LIGHT : FORWARD_ID_TYPE_SPOT_LIGHT, li->forward_id, index); + } + Transform3D light_transform = li->transform; float sign = storage->light_is_negative(base) ? -1 : 1; @@ -2695,22 +3211,19 @@ void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const light_data.shadow_enabled = true; - if (type == RS::LIGHT_SPOT) { - light_data.shadow_bias = (storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_BIAS) * radius / 10.0); - float shadow_texel_size = Math::tan(Math::deg2rad(spot_angle)) * radius * 2.0; - shadow_texel_size *= light_instance_get_shadow_texel_size(li->self, p_shadow_atlas); - - light_data.shadow_normal_bias = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_NORMAL_BIAS) * shadow_texel_size; + float shadow_texel_size = light_instance_get_shadow_texel_size(li->self, p_shadow_atlas); + light_data.shadow_normal_bias = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_NORMAL_BIAS) * shadow_texel_size * 10.0; + if (type == RS::LIGHT_SPOT) { + light_data.shadow_bias = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_BIAS) / 100.0; } else { //omni - light_data.shadow_bias = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_BIAS) * radius / 10.0; - float shadow_texel_size = light_instance_get_shadow_texel_size(li->self, p_shadow_atlas); - light_data.shadow_normal_bias = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_NORMAL_BIAS) * shadow_texel_size * 2.0; // applied in -1 .. 1 space + light_data.shadow_bias = storage->light_get_param(base, RS::LIGHT_PARAM_SHADOW_BIAS); } light_data.transmittance_bias = storage->light_get_transmittance_bias(base); - Rect2 rect = light_instance_get_shadow_atlas_rect(li->self, p_shadow_atlas); + Vector2i omni_offset; + Rect2 rect = light_instance_get_shadow_atlas_rect(li->self, p_shadow_atlas, omni_offset); light_data.atlas_rect[0] = rect.position.x; light_data.atlas_rect[1] = rect.position.y; @@ -2721,7 +3234,6 @@ void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const light_data.shadow_volumetric_fog_fade = 1.0 / storage->light_get_shadow_volumetric_fog_fade(base); if (type == RS::LIGHT_OMNI) { - light_data.atlas_rect[3] *= 0.5; //one paraboloid on top of another Transform3D proj = (inverse_transform * light_transform).inverse(); RendererStorageRD::store_transform(proj, light_data.shadow_matrix); @@ -2733,6 +3245,8 @@ void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const light_data.soft_shadow_scale *= shadows_quality_radius_get(); // Only use quality radius for PCF } + light_data.direction[0] = omni_offset.x * float(rect.size.width); + light_data.direction[1] = omni_offset.y * float(rect.size.height); } else if (type == RS::LIGHT_SPOT) { Transform3D modelview = (inverse_transform * light_transform).inverse(); CameraMatrix bias; @@ -2754,7 +3268,6 @@ void RendererSceneRenderRD::_setup_lights(const PagedArray<RID> &p_lights, const light_data.shadow_enabled = false; } - li->light_index = index; li->cull_mask = storage->light_get_cull_mask(base); if (current_cluster_builder != nullptr) { @@ -2823,11 +3336,15 @@ void RendererSceneRenderRD::_setup_decals(const PagedArray<RID> &p_decals, const sort_array.sort(cluster.decal_sort, cluster.decal_count); } + bool using_forward_ids = _uses_forward_ids(); for (uint32_t i = 0; i < cluster.decal_count; i++) { DecalInstance *di = cluster.decal_sort[i].instance; RID decal = di->decal; - di->render_index = i; + if (using_forward_ids) { + _map_forward_id(FORWARD_ID_TYPE_DECAL, di->forward_id, i); + } + di->cull_mask = storage->decal_get_cull_mask(decal); Transform3D xform = di->transform; @@ -2944,116 +3461,6 @@ void RendererSceneRenderRD::_setup_decals(const PagedArray<RID> &p_decals, const } } -void RendererSceneRenderRD::_fill_instance_indices(const RID *p_omni_light_instances, uint32_t p_omni_light_instance_count, uint32_t *p_omni_light_indices, const RID *p_spot_light_instances, uint32_t p_spot_light_instance_count, uint32_t *p_spot_light_indices, const RID *p_reflection_probe_instances, uint32_t p_reflection_probe_instance_count, uint32_t *p_reflection_probe_indices, const RID *p_decal_instances, uint32_t p_decal_instance_count, uint32_t *p_decal_instance_indices, uint32_t p_layer_mask, uint32_t p_max_dst_words) { - // first zero out our indices - for (uint32_t i = 0; i < p_max_dst_words; i++) { - p_omni_light_indices[i] = 0; - p_spot_light_indices[i] = 0; - p_reflection_probe_indices[i] = 0; - p_decal_instance_indices[i] = 0; - } - - { - // process omni lights - uint32_t dword = 0; - uint32_t shift = 0; - - for (uint32_t i = 0; i < p_omni_light_instance_count && dword < p_max_dst_words; i++) { - LightInstance *li = light_instance_owner.getornull(p_omni_light_instances[i]); - - if ((li->cull_mask & p_layer_mask) && (li->light_index < 255)) { - p_omni_light_indices[dword] += li->light_index << shift; - if (shift == 24) { - dword++; - shift = 0; - } else { - shift += 8; - } - } - } - - if (dword < 2) { - // put in ending mark - p_omni_light_indices[dword] += 0xFF << shift; - } - } - - { - // process spot lights - uint32_t dword = 0; - uint32_t shift = 0; - - for (uint32_t i = 0; i < p_spot_light_instance_count && dword < p_max_dst_words; i++) { - LightInstance *li = light_instance_owner.getornull(p_spot_light_instances[i]); - - if ((li->cull_mask & p_layer_mask) && (li->light_index < 255)) { - p_spot_light_indices[dword] += li->light_index << shift; - if (shift == 24) { - dword++; - shift = 0; - } else { - shift += 8; - } - } - } - - if (dword < 2) { - // put in ending mark - p_spot_light_indices[dword] += 0xFF << shift; - } - } - - { - // process reflection probes - uint32_t dword = 0; - uint32_t shift = 0; - - for (uint32_t i = 0; i < p_reflection_probe_instance_count && dword < p_max_dst_words; i++) { - ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(p_reflection_probe_instances[i]); - - if ((rpi->cull_mask & p_layer_mask) && (rpi->render_index < 255)) { - p_reflection_probe_indices[dword] += rpi->render_index << shift; - if (shift == 24) { - dword++; - shift = 0; - } else { - shift += 8; - } - } - } - - if (dword < 2) { - // put in ending mark - p_reflection_probe_indices[dword] += 0xFF << shift; - } - } - - { - // process decals - uint32_t dword = 0; - uint32_t shift = 0; - - for (uint32_t i = 0; i < p_decal_instance_count && dword < p_max_dst_words; i++) { - DecalInstance *decal = decal_instance_owner.getornull(p_decal_instances[i]); - - if ((decal->cull_mask & p_layer_mask) && (decal->render_index < 255)) { - p_decal_instance_indices[dword] += decal->render_index << shift; - if (shift == 24) { - dword++; - shift = 0; - } else { - shift += 8; - } - } - } - - if (dword < 2) { - // put in ending mark - p_decal_instance_indices[dword] += 0xFF << shift; - } - } -} - void RendererSceneRenderRD::_volumetric_fog_erase(RenderBuffers *rb) { ERR_FAIL_COND(!rb->volumetric_fog); @@ -3522,7 +3929,6 @@ void RendererSceneRenderRD::_pre_opaque_render(RenderDataRD *p_render_data, bool Plane camera_plane(p_render_data->cam_transform.origin, -p_render_data->cam_transform.basis.get_axis(Vector3::AXIS_Z)); float lod_distance_multiplier = p_render_data->cam_projection.get_lod_multiplier(); - { for (int i = 0; i < render_state.render_shadow_count; i++) { LightInstance *li = light_instance_owner.getornull(render_state.render_shadows[i].light); @@ -3538,7 +3944,7 @@ void RendererSceneRenderRD::_pre_opaque_render(RenderDataRD *p_render_data, bool //cube shadows are rendered in their own way for (uint32_t i = 0; i < render_state.cube_shadows.size(); i++) { - _render_shadow_pass(render_state.render_shadows[render_state.cube_shadows[i]].light, p_render_data->shadow_atlas, render_state.render_shadows[render_state.cube_shadows[i]].pass, render_state.render_shadows[render_state.cube_shadows[i]].instances, camera_plane, lod_distance_multiplier, p_render_data->screen_lod_threshold, true, true, true); + _render_shadow_pass(render_state.render_shadows[render_state.cube_shadows[i]].light, p_render_data->shadow_atlas, render_state.render_shadows[render_state.cube_shadows[i]].pass, render_state.render_shadows[render_state.cube_shadows[i]].instances, camera_plane, lod_distance_multiplier, p_render_data->screen_lod_threshold, true, true, true, p_render_data->render_info); } if (render_state.directional_shadows.size()) { @@ -3568,11 +3974,11 @@ void RendererSceneRenderRD::_pre_opaque_render(RenderDataRD *p_render_data, bool //render directional shadows for (uint32_t i = 0; i < render_state.directional_shadows.size(); i++) { - _render_shadow_pass(render_state.render_shadows[render_state.directional_shadows[i]].light, p_render_data->shadow_atlas, render_state.render_shadows[render_state.directional_shadows[i]].pass, render_state.render_shadows[render_state.directional_shadows[i]].instances, camera_plane, lod_distance_multiplier, p_render_data->screen_lod_threshold, false, i == render_state.directional_shadows.size() - 1, false); + _render_shadow_pass(render_state.render_shadows[render_state.directional_shadows[i]].light, p_render_data->shadow_atlas, render_state.render_shadows[render_state.directional_shadows[i]].pass, render_state.render_shadows[render_state.directional_shadows[i]].instances, camera_plane, lod_distance_multiplier, p_render_data->screen_lod_threshold, false, i == render_state.directional_shadows.size() - 1, false, p_render_data->render_info); } //render positional shadows for (uint32_t i = 0; i < render_state.shadows.size(); i++) { - _render_shadow_pass(render_state.render_shadows[render_state.shadows[i]].light, p_render_data->shadow_atlas, render_state.render_shadows[render_state.shadows[i]].pass, render_state.render_shadows[render_state.shadows[i]].instances, camera_plane, lod_distance_multiplier, p_render_data->screen_lod_threshold, i == 0, i == render_state.shadows.size() - 1, true); + _render_shadow_pass(render_state.render_shadows[render_state.shadows[i]].light, p_render_data->shadow_atlas, render_state.render_shadows[render_state.shadows[i]].pass, render_state.render_shadows[render_state.shadows[i]].instances, camera_plane, lod_distance_multiplier, p_render_data->screen_lod_threshold, i == 0, i == render_state.shadows.size() - 1, true, p_render_data->render_info); } _render_shadow_process(); @@ -3618,7 +4024,7 @@ void RendererSceneRenderRD::_pre_opaque_render(RenderDataRD *p_render_data, bool uint32_t directional_light_count = 0; uint32_t positional_light_count = 0; - _setup_lights(*p_render_data->lights, p_render_data->cam_transform, p_render_data->shadow_atlas, using_shadows, directional_light_count, positional_light_count); + _setup_lights(*p_render_data->lights, p_render_data->cam_transform, p_render_data->shadow_atlas, using_shadows, directional_light_count, positional_light_count, p_render_data->directional_light_soft_shadows); _setup_decals(*p_render_data->decals, p_render_data->cam_transform.affine_inverse()); p_render_data->directional_light_count = directional_light_count; @@ -3641,7 +4047,7 @@ void RendererSceneRenderRD::_pre_opaque_render(RenderDataRD *p_render_data, bool } } -void RendererSceneRenderRD::render_scene(RID p_render_buffers, const CameraData *p_camera_data, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_lights, const PagedArray<RID> &p_reflection_probes, const PagedArray<RID> &p_voxel_gi_instances, const PagedArray<RID> &p_decals, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_occluder_debug_tex, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold, const RenderShadowData *p_render_shadows, int p_render_shadow_count, const RenderSDFGIData *p_render_sdfgi_regions, int p_render_sdfgi_region_count, const RenderSDFGIUpdateData *p_sdfgi_update_data) { +void RendererSceneRenderRD::render_scene(RID p_render_buffers, const CameraData *p_camera_data, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_lights, const PagedArray<RID> &p_reflection_probes, const PagedArray<RID> &p_voxel_gi_instances, const PagedArray<RID> &p_decals, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_occluder_debug_tex, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold, const RenderShadowData *p_render_shadows, int p_render_shadow_count, const RenderSDFGIData *p_render_sdfgi_regions, int p_render_sdfgi_region_count, const RenderSDFGIUpdateData *p_sdfgi_update_data, RendererScene::RenderInfo *r_render_info) { // getting this here now so we can direct call a bunch of things more easily RenderBuffers *rb = nullptr; if (p_render_buffers.is_valid()) { @@ -3696,6 +4102,7 @@ void RendererSceneRenderRD::render_scene(RID p_render_buffers, const CameraData render_state.render_sdfgi_regions = p_render_sdfgi_regions; render_state.render_sdfgi_region_count = p_render_sdfgi_region_count; render_state.sdfgi_update_data = p_sdfgi_update_data; + render_data.render_info = r_render_info; } PagedArray<RID> empty; @@ -3752,16 +4159,14 @@ void RendererSceneRenderRD::render_scene(RID p_render_buffers, const CameraData render_state.voxel_gi_count = 0; - if (rb != nullptr) { + if (rb != nullptr && is_dynamic_gi_supported()) { if (rb->sdfgi) { rb->sdfgi->update_cascades(); rb->sdfgi->pre_process_gi(render_data.cam_transform, &render_data, this); rb->sdfgi->update_light(); } - if (p_voxel_gi_instances.size()) { - gi.setup_voxel_gi_instances(render_data.render_buffers, render_data.cam_transform, *render_data.voxel_gi_instances, render_state.voxel_gi_count, this); - } + gi.setup_voxel_gi_instances(render_data.render_buffers, render_data.cam_transform, *render_data.voxel_gi_instances, render_state.voxel_gi_count, this); } render_state.depth_prepass_used = false; @@ -3775,9 +4180,28 @@ void RendererSceneRenderRD::render_scene(RID p_render_buffers, const CameraData _render_scene(&render_data, clear_color); if (p_render_buffers.is_valid()) { - if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_OMNI_LIGHTS || debug_draw == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_SPOT_LIGHTS || debug_draw == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_DECALS || debug_draw == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_REFLECTION_PROBES) { + /* + _debug_draw_cluster(p_render_buffers); + + RENDER_TIMESTAMP("Tonemap"); + + _render_buffers_post_process_and_tonemap(&render_data); + */ + + _render_buffers_debug_draw(p_render_buffers, p_shadow_atlas, p_occluder_debug_tex); + if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_SDFGI && rb != nullptr && rb->sdfgi != nullptr) { + rb->sdfgi->debug_draw(render_data.cam_projection, render_data.cam_transform, rb->width, rb->height, rb->render_target, rb->texture); + } + } +} + +void RendererSceneRenderRD::_debug_draw_cluster(RID p_render_buffers) { + if (p_render_buffers.is_valid() && current_cluster_builder != nullptr) { + RS::ViewportDebugDraw dd = get_debug_draw_mode(); + + if (dd == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_OMNI_LIGHTS || dd == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_SPOT_LIGHTS || dd == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_DECALS || dd == RS::VIEWPORT_DEBUG_DRAW_CLUSTER_REFLECTION_PROBES) { ClusterBuilderRD::ElementType elem_type = ClusterBuilderRD::ELEMENT_TYPE_MAX; - switch (debug_draw) { + switch (dd) { case RS::VIEWPORT_DEBUG_DRAW_CLUSTER_OMNI_LIGHTS: elem_type = ClusterBuilderRD::ELEMENT_TYPE_OMNI_LIGHT; break; @@ -3793,22 +4217,12 @@ void RendererSceneRenderRD::render_scene(RID p_render_buffers, const CameraData default: { } } - if (current_cluster_builder != nullptr) { - current_cluster_builder->debug(elem_type); - } - } - - RENDER_TIMESTAMP("Tonemap"); - - _render_buffers_post_process_and_tonemap(&render_data); - _render_buffers_debug_draw(p_render_buffers, p_shadow_atlas, p_occluder_debug_tex); - if (debug_draw == RS::VIEWPORT_DEBUG_DRAW_SDFGI && rb != nullptr && rb->sdfgi != nullptr) { - rb->sdfgi->debug_draw(render_data.cam_projection, render_data.cam_transform, rb->width, rb->height, rb->render_target, rb->texture); + current_cluster_builder->debug(elem_type); } } } -void RendererSceneRenderRD::_render_shadow_pass(RID p_light, RID p_shadow_atlas, int p_pass, const PagedArray<GeometryInstance *> &p_instances, const Plane &p_camera_plane, float p_lod_distance_multiplier, float p_screen_lod_threshold, bool p_open_pass, bool p_close_pass, bool p_clear_region) { +void RendererSceneRenderRD::_render_shadow_pass(RID p_light, RID p_shadow_atlas, int p_pass, const PagedArray<GeometryInstance *> &p_instances, const Plane &p_camera_plane, float p_lod_distance_multiplier, float p_screen_lod_threshold, bool p_open_pass, bool p_close_pass, bool p_clear_region, RendererScene::RenderInfo *p_render_info) { LightInstance *light_instance = light_instance_owner.getornull(p_light); ERR_FAIL_COND(!light_instance); @@ -3818,6 +4232,7 @@ void RendererSceneRenderRD::_render_shadow_pass(RID p_light, RID p_shadow_atlas, bool using_dual_paraboloid = false; bool using_dual_paraboloid_flip = false; + Vector2i dual_paraboloid_offset; RID render_fb; RID render_texture; float zfar; @@ -3911,6 +4326,9 @@ void RendererSceneRenderRD::_render_shadow_pass(RID p_light, RID p_shadow_atlas, zfar = storage->light_get_param(light_instance->light, RS::LIGHT_PARAM_RANGE); if (storage->light_get_type(light_instance->light) == RS::LIGHT_OMNI) { + bool wrap = (shadow + 1) % shadow_atlas->quadrants[quadrant].subdivision == 0; + dual_paraboloid_offset = wrap ? Vector2i(1 - shadow_atlas->quadrants[quadrant].subdivision, 1) : Vector2i(1, 0); + if (storage->light_omni_get_shadow_mode(light_instance->light) == RS::LIGHT_OMNI_SHADOW_CUBE) { ShadowCubemap *cubemap = _get_shadow_cubemap(shadow_size / 2); @@ -3930,12 +4348,16 @@ void RendererSceneRenderRD::_render_shadow_pass(RID p_light, RID p_shadow_atlas, } } else { + atlas_rect.position.x += 1; + atlas_rect.position.y += 1; + atlas_rect.size.x -= 2; + atlas_rect.size.y -= 2; + + atlas_rect.position += p_pass * atlas_rect.size * dual_paraboloid_offset; + light_projection = light_instance->shadow_transform[0].camera; light_transform = light_instance->shadow_transform[0].transform; - atlas_rect.size.height /= 2; - atlas_rect.position.y += p_pass * atlas_rect.size.height; - using_dual_paraboloid = true; using_dual_paraboloid_flip = p_pass == 1; render_fb = shadow_atlas->fb; @@ -3954,7 +4376,7 @@ void RendererSceneRenderRD::_render_shadow_pass(RID p_light, RID p_shadow_atlas, if (render_cubemap) { //rendering to cubemap - _render_shadow_append(render_fb, p_instances, light_projection, light_transform, zfar, 0, 0, false, false, use_pancake, p_camera_plane, p_lod_distance_multiplier, p_screen_lod_threshold, Rect2(), false, true, true, true); + _render_shadow_append(render_fb, p_instances, light_projection, light_transform, zfar, 0, 0, false, false, use_pancake, p_camera_plane, p_lod_distance_multiplier, p_screen_lod_threshold, Rect2(), false, true, true, true, p_render_info); if (finalize_cubemap) { _render_shadow_process(); _render_shadow_end(); @@ -3964,10 +4386,9 @@ void RendererSceneRenderRD::_render_shadow_pass(RID p_light, RID p_shadow_atlas, atlas_rect_norm.position.y /= float(atlas_size); atlas_rect_norm.size.x /= float(atlas_size); atlas_rect_norm.size.y /= float(atlas_size); - atlas_rect_norm.size.height /= 2; - storage->get_effects()->copy_cubemap_to_dp(render_texture, atlas_fb, atlas_rect_norm, light_projection.get_z_near(), light_projection.get_z_far(), false); - atlas_rect_norm.position.y += atlas_rect_norm.size.height; - storage->get_effects()->copy_cubemap_to_dp(render_texture, atlas_fb, atlas_rect_norm, light_projection.get_z_near(), light_projection.get_z_far(), true); + storage->get_effects()->copy_cubemap_to_dp(render_texture, atlas_fb, atlas_rect_norm, atlas_rect.size, light_projection.get_z_near(), light_projection.get_z_far(), false); + atlas_rect_norm.position += Vector2(dual_paraboloid_offset) * atlas_rect_norm.size; + storage->get_effects()->copy_cubemap_to_dp(render_texture, atlas_fb, atlas_rect_norm, atlas_rect.size, light_projection.get_z_near(), light_projection.get_z_far(), true); //restore transform so it can be properly used light_instance_set_shadow_transform(p_light, CameraMatrix(), light_instance->transform, zfar, 0, 0, 0); @@ -3975,7 +4396,7 @@ void RendererSceneRenderRD::_render_shadow_pass(RID p_light, RID p_shadow_atlas, } else { //render shadow - _render_shadow_append(render_fb, p_instances, light_projection, light_transform, zfar, 0, 0, using_dual_paraboloid, using_dual_paraboloid_flip, use_pancake, p_camera_plane, p_lod_distance_multiplier, p_screen_lod_threshold, atlas_rect, flip_y, p_clear_region, p_open_pass, p_close_pass); + _render_shadow_append(render_fb, p_instances, light_projection, light_transform, zfar, 0, 0, using_dual_paraboloid, using_dual_paraboloid_flip, use_pancake, p_camera_plane, p_lod_distance_multiplier, p_screen_lod_threshold, atlas_rect, flip_y, p_clear_region, p_open_pass, p_close_pass, p_render_info); } } @@ -4031,11 +4452,13 @@ bool RendererSceneRenderRD::free(RID p_rid) { } reflection_atlas_owner.free(p_rid); } else if (reflection_probe_instance_owner.owns(p_rid)) { - //not much to delete, just free it - //ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(p_rid); + ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(p_rid); + _free_forward_id(FORWARD_ID_TYPE_REFLECTION_PROBE, rpi->forward_id); reflection_probe_release_atlas_index(p_rid); reflection_probe_instance_owner.free(p_rid); } else if (decal_instance_owner.owns(p_rid)) { + DecalInstance *di = decal_instance_owner.getornull(p_rid); + _free_forward_id(FORWARD_ID_TYPE_DECAL, di->forward_id); decal_instance_owner.free(p_rid); } else if (lightmap_instance_owner.owns(p_rid)) { lightmap_instance_owner.free(p_rid); @@ -4067,9 +4490,18 @@ bool RendererSceneRenderRD::free(RID p_rid) { uint32_t s = key & ShadowAtlas::SHADOW_INDEX_MASK; shadow_atlas->quadrants[q].shadows.write[s].owner = RID(); + + if (key & ShadowAtlas::OMNI_LIGHT_FLAG) { + // Omni lights use two atlas spots, make sure to clear the other as well + shadow_atlas->quadrants[q].shadows.write[s + 1].owner = RID(); + } + shadow_atlas->shadow_owners.erase(p_rid); } + if (light_instance->light_type != RS::LIGHT_DIRECTIONAL) { + _free_forward_id(light_instance->light_type == RS::LIGHT_OMNI ? FORWARD_ID_TYPE_OMNI_LIGHT : FORWARD_ID_TYPE_SPOT_LIGHT, light_instance->forward_id); + } light_instance_owner.free(p_rid); } else if (shadow_atlas_owner.owns(p_rid)) { @@ -4174,7 +4606,7 @@ TypedArray<Image> RendererSceneRenderRD::bake_render_uv2(RID p_base, const Vecto { PackedByteArray data = RD::get_singleton()->texture_get_data(albedo_alpha_tex, 0); Ref<Image> img; - img.instance(); + img.instantiate(); img->create(p_image_size.width, p_image_size.height, false, Image::FORMAT_RGBA8, data); RD::get_singleton()->free(albedo_alpha_tex); ret.push_back(img); @@ -4183,7 +4615,7 @@ TypedArray<Image> RendererSceneRenderRD::bake_render_uv2(RID p_base, const Vecto { PackedByteArray data = RD::get_singleton()->texture_get_data(normal_tex, 0); Ref<Image> img; - img.instance(); + img.instantiate(); img->create(p_image_size.width, p_image_size.height, false, Image::FORMAT_RGBA8, data); RD::get_singleton()->free(normal_tex); ret.push_back(img); @@ -4192,7 +4624,7 @@ TypedArray<Image> RendererSceneRenderRD::bake_render_uv2(RID p_base, const Vecto { PackedByteArray data = RD::get_singleton()->texture_get_data(orm_tex, 0); Ref<Image> img; - img.instance(); + img.instantiate(); img->create(p_image_size.width, p_image_size.height, false, Image::FORMAT_RGBA8, data); RD::get_singleton()->free(orm_tex); ret.push_back(img); @@ -4201,7 +4633,7 @@ TypedArray<Image> RendererSceneRenderRD::bake_render_uv2(RID p_base, const Vecto { PackedByteArray data = RD::get_singleton()->texture_get_data(emission_tex, 0); Ref<Image> img; - img.instance(); + img.instantiate(); img->create(p_image_size.width, p_image_size.height, false, Image::FORMAT_RGBAH, data); RD::get_singleton()->free(emission_tex); ret.push_back(img); @@ -4261,10 +4693,12 @@ uint32_t RendererSceneRenderRD::get_max_elements() const { } RendererSceneRenderRD::RendererSceneRenderRD(RendererStorageRD *p_storage) { - max_cluster_elements = get_max_elements(); - storage = p_storage; singleton = this; +} + +void RendererSceneRenderRD::init() { + max_cluster_elements = get_max_elements(); directional_shadow.size = GLOBAL_GET("rendering/shadows/directional_shadow/size"); directional_shadow.use_16_bits = GLOBAL_GET("rendering/shadows/directional_shadow/16_bits"); @@ -4359,6 +4793,9 @@ RendererSceneRenderRD::RendererSceneRenderRD(RendererStorageRD *p_storage) { environment_set_volumetric_fog_volume_size(GLOBAL_GET("rendering/environment/volumetric_fog/volume_size"), GLOBAL_GET("rendering/environment/volumetric_fog/volume_depth")); environment_set_volumetric_fog_filter_active(GLOBAL_GET("rendering/environment/volumetric_fog/use_filter")); + decals_set_filter(RS::DecalFilter(int(GLOBAL_GET("rendering/textures/decals/filter")))); + light_projectors_set_filter(RS::LightProjectorFilter(int(GLOBAL_GET("rendering/textures/light_projectors/filter")))); + cull_argument.set_page_pool(&cull_argument_pool); } diff --git a/servers/rendering/renderer_rd/renderer_scene_render_rd.h b/servers/rendering/renderer_rd/renderer_scene_render_rd.h index 3096881afa..eb61af517a 100644 --- a/servers/rendering/renderer_rd/renderer_scene_render_rd.h +++ b/servers/rendering/renderer_rd/renderer_scene_render_rd.h @@ -40,6 +40,7 @@ #include "servers/rendering/renderer_rd/renderer_scene_sky_rd.h" #include "servers/rendering/renderer_rd/renderer_storage_rd.h" #include "servers/rendering/renderer_rd/shaders/volumetric_fog.glsl.gen.h" +#include "servers/rendering/renderer_scene.h" #include "servers/rendering/renderer_scene_render.h" #include "servers/rendering/rendering_device.h" @@ -79,6 +80,9 @@ struct RenderDataRD { uint32_t cluster_max_elements = 0; uint32_t directional_light_count = 0; + bool directional_light_soft_shadows = false; + + RendererScene::RenderInfo *render_info = nullptr; }; class RendererSceneRenderRD : public RendererSceneRender { @@ -91,19 +95,19 @@ protected: double time_step = 0; struct RenderBufferData { - virtual void configure(RID p_color_buffer, RID p_depth_buffer, int p_width, int p_height, RS::ViewportMSAA p_msaa, uint32_t p_view_count) = 0; + virtual void configure(RID p_color_buffer, RID p_depth_buffer, RID p_target_buffer, int p_width, int p_height, RS::ViewportMSAA p_msaa, uint32_t p_view_count) = 0; virtual ~RenderBufferData() {} }; virtual RenderBufferData *_create_render_buffer_data() = 0; - void _setup_lights(const PagedArray<RID> &p_lights, const Transform3D &p_camera_transform, RID p_shadow_atlas, bool p_using_shadows, uint32_t &r_directional_light_count, uint32_t &r_positional_light_count); + void _setup_lights(const PagedArray<RID> &p_lights, const Transform3D &p_camera_transform, RID p_shadow_atlas, bool p_using_shadows, uint32_t &r_directional_light_count, uint32_t &r_positional_light_count, bool &r_directional_light_soft_shadows); void _setup_decals(const PagedArray<RID> &p_decals, const Transform3D &p_camera_inverse_xform); void _setup_reflections(const PagedArray<RID> &p_reflections, const Transform3D &p_camera_inverse_transform, RID p_environment); virtual void _render_scene(RenderDataRD *p_render_data, const Color &p_default_color) = 0; virtual void _render_shadow_begin() = 0; - virtual void _render_shadow_append(RID p_framebuffer, const PagedArray<GeometryInstance *> &p_instances, const CameraMatrix &p_projection, const Transform3D &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0, const Rect2i &p_rect = Rect2i(), bool p_flip_y = false, bool p_clear_region = true, bool p_begin = true, bool p_end = true) = 0; + virtual void _render_shadow_append(RID p_framebuffer, const PagedArray<GeometryInstance *> &p_instances, const CameraMatrix &p_projection, const Transform3D &p_transform, float p_zfar, float p_bias, float p_normal_bias, bool p_use_dp, bool p_use_dp_flip, bool p_use_pancake, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0.0, float p_screen_lod_threshold = 0.0, const Rect2i &p_rect = Rect2i(), bool p_flip_y = false, bool p_clear_region = true, bool p_begin = true, bool p_end = true, RendererScene::RenderInfo *p_render_info = nullptr) = 0; virtual void _render_shadow_process() = 0; virtual void _render_shadow_end(uint32_t p_barrier = RD::BARRIER_MASK_ALL) = 0; @@ -113,6 +117,7 @@ protected: virtual void _render_particle_collider_heightfield(RID p_fb, const Transform3D &p_cam_transform, const CameraMatrix &p_cam_projection, const PagedArray<GeometryInstance *> &p_instances) = 0; void _debug_sdfgi_probes(RID p_render_buffers, RD::DrawListID p_draw_list, RID p_framebuffer, const CameraMatrix &p_camera_with_transform); + void _debug_draw_cluster(RID p_render_buffers); RenderBufferData *render_buffers_get_data(RID p_render_buffers); @@ -129,6 +134,12 @@ protected: void _pre_opaque_render(RenderDataRD *p_render_data, bool p_use_ssao, bool p_use_gi, RID p_normal_roughness_buffer, RID p_voxel_gi_buffer); + void _render_buffers_copy_screen_texture(const RenderDataRD *p_render_data); + void _render_buffers_copy_depth_texture(const RenderDataRD *p_render_data); + void _render_buffers_post_process_and_tonemap(const RenderDataRD *p_render_data); + void _post_process_subpass(RID p_source_texture, RID p_framebuffer, const RenderDataRD *p_render_data); + void _disable_clear_request(const RenderDataRD *p_render_data); + // needed for a single argument calls (material and uv2) PagedArrayPool<GeometryInstance *> cull_argument_pool; PagedArray<GeometryInstance *> cull_argument; //need this to exist @@ -142,7 +153,26 @@ protected: } else { return nullptr; } - } + }; + + //used for mobile renderer mostly + + typedef int32_t ForwardID; + + enum ForwardIDType { + FORWARD_ID_TYPE_OMNI_LIGHT, + FORWARD_ID_TYPE_SPOT_LIGHT, + FORWARD_ID_TYPE_REFLECTION_PROBE, + FORWARD_ID_TYPE_DECAL, + FORWARD_ID_MAX, + }; + + virtual ForwardID _allocate_forward_id(ForwardIDType p_type) { return -1; } + virtual void _free_forward_id(ForwardIDType p_type, ForwardID p_id) {} + virtual void _map_forward_id(ForwardIDType p_type, ForwardID p_id, uint32_t p_index) {} + virtual bool _uses_forward_ids() const { return false; } + + virtual void _update_shader_quality_settings() {} private: RS::ViewportDebugDraw debug_draw = RS::VIEWPORT_DEBUG_DRAW_DISABLED; @@ -185,9 +215,10 @@ private: uint32_t render_step = 0; uint64_t last_pass = 0; - uint32_t render_index = 0; uint32_t cull_mask = 0; + ForwardID forward_id = -1; + Transform3D transform; }; @@ -198,8 +229,8 @@ private: struct DecalInstance { RID decal; Transform3D transform; - uint32_t render_index; uint32_t cull_mask; + ForwardID forward_id = -1; }; mutable RID_Owner<DecalInstance> decal_instance_owner; @@ -224,7 +255,8 @@ private: struct ShadowAtlas { enum { QUADRANT_SHIFT = 27, - SHADOW_INDEX_MASK = (1 << QUADRANT_SHIFT) - 1, + OMNI_LIGHT_FLAG = 1 << 26, + SHADOW_INDEX_MASK = OMNI_LIGHT_FLAG - 1, SHADOW_INVALID = 0xFFFFFFFF }; @@ -268,7 +300,9 @@ private: void _update_shadow_atlas(ShadowAtlas *shadow_atlas); + void _shadow_atlas_invalidate_shadow(RendererSceneRenderRD::ShadowAtlas::Quadrant::Shadow *p_shadow, RID p_atlas, RendererSceneRenderRD::ShadowAtlas *p_shadow_atlas, uint32_t p_quadrant, uint32_t p_shadow_idx); bool _shadow_atlas_find_shadow(ShadowAtlas *shadow_atlas, int *p_in_quadrants, int p_quadrant_count, int p_current_subdiv, uint64_t p_tick, int &r_quadrant, int &r_shadow); + bool _shadow_atlas_find_omni_shadows(ShadowAtlas *shadow_atlas, int *p_in_quadrants, int p_quadrant_count, int p_current_subdiv, uint64_t p_tick, int &r_quadrant, int &r_shadow); RS::ShadowQuality shadows_quality = RS::SHADOW_QUALITY_MAX; //So it always updates when first set RS::ShadowQuality directional_shadow_quality = RS::SHADOW_QUALITY_MAX; @@ -283,6 +317,8 @@ private: int directional_soft_shadow_samples = 0; int penumbra_shadow_samples = 0; int soft_shadow_samples = 0; + RS::DecalFilter decals_filter = RS::DECAL_FILTER_LINEAR_MIPMAPS; + RS::LightProjectorFilter light_projectors_filter = RS::LIGHT_PROJECTOR_FILTER_LINEAR_MIPMAPS; /* DIRECTIONAL SHADOW */ @@ -343,18 +379,15 @@ private: uint64_t last_scene_pass = 0; uint64_t last_scene_shadow_pass = 0; uint64_t last_pass = 0; - uint32_t light_index = 0; uint32_t cull_mask = 0; uint32_t light_directional_index = 0; - uint32_t current_shadow_atlas_key = 0; - - Vector2 dp; - Rect2 directional_rect; Set<RID> shadow_atlases; //shadow atlases where this light is registered + ForwardID forward_id = -1; + LightInstance() {} }; @@ -423,6 +456,7 @@ private: RID texture; //main texture for rendering to, must be filled after done rendering RID depth_texture; //main depth texture + RID texture_fb; // framebuffer for the main texture, ONLY USED FOR MOBILE RENDERER POST EFFECTS, DO NOT USE FOR RENDERING 3D!!! RendererSceneGIRD::SDFGI *sdfgi = nullptr; VolumetricFog *volumetric_fog = nullptr; @@ -438,6 +472,11 @@ private: RID texture; int width; int height; + + // only used on mobile renderer + RID fb; + RID half_texture; + RID half_fb; }; Vector<Mipmap> mipmaps; @@ -445,9 +484,25 @@ private: Blur blur[2]; //the second one starts from the first mipmap + struct WeightBuffers { + RID weight; + RID fb; // FB with both texture and weight + }; + + // 2 full size, 2 half size + WeightBuffers weight_buffers[4]; // Only used in raster + RID base_weight_fb; // base buffer for weight + + RID depth_back_texture; + RID depth_back_fb; // only used on mobile + struct Luminance { Vector<RID> reduce; RID current; + + // used only on mobile renderer + Vector<RID> fb; + RID current_fb; } luminance; struct SSAO { @@ -459,6 +514,10 @@ private: Vector<RID> ao_pong_slices; RID ao_final; RID importance_map[2]; + + RID downsample_uniform_set; + RID gather_uniform_set; + RID importance_map_uniform_set; } ssao; struct SSR { @@ -480,10 +539,10 @@ private: void _free_render_buffer_data(RenderBuffers *rb); void _allocate_blur_textures(RenderBuffers *rb); + void _allocate_depth_backbuffer_textures(RenderBuffers *rb); void _allocate_luminance_textures(RenderBuffers *rb); void _render_buffers_debug_draw(RID p_render_buffers, RID p_shadow_atlas, RID p_occlusion_buffer); - void _render_buffers_post_process_and_tonemap(const RenderDataRD *p_render_data); /* Cluster */ @@ -515,7 +574,7 @@ private: struct LightData { float position[3]; float inv_radius; - float direction[3]; + float direction[3]; // in omni, x and y are used for dual paraboloid offset float size; float color[3]; @@ -742,7 +801,7 @@ private: uint32_t max_cluster_elements = 512; - void _render_shadow_pass(RID p_light, RID p_shadow_atlas, int p_pass, const PagedArray<GeometryInstance *> &p_instances, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0, float p_screen_lod_threshold = 0.0, bool p_open_pass = true, bool p_close_pass = true, bool p_clear_region = true); + void _render_shadow_pass(RID p_light, RID p_shadow_atlas, int p_pass, const PagedArray<GeometryInstance *> &p_instances, const Plane &p_camera_plane = Plane(), float p_lod_distance_multiplier = 0, float p_screen_lod_threshold = 0.0, bool p_open_pass = true, bool p_close_pass = true, bool p_clear_region = true, RendererScene::RenderInfo *p_render_info = nullptr); public: virtual Transform3D geometry_instance_get_transform(GeometryInstance *p_instance) = 0; @@ -750,10 +809,10 @@ public: /* SHADOW ATLAS API */ - RID shadow_atlas_create(); - void shadow_atlas_set_size(RID p_atlas, int p_size, bool p_16_bits = false); - void shadow_atlas_set_quadrant_subdivision(RID p_atlas, int p_quadrant, int p_subdivision); - bool shadow_atlas_update_light(RID p_atlas, RID p_light_intance, float p_coverage, uint64_t p_light_version); + virtual RID shadow_atlas_create() override; + virtual void shadow_atlas_set_size(RID p_atlas, int p_size, bool p_16_bits = false) override; + virtual void shadow_atlas_set_quadrant_subdivision(RID p_atlas, int p_quadrant, int p_subdivision) override; + virtual bool shadow_atlas_update_light(RID p_atlas, RID p_light_intance, float p_coverage, uint64_t p_light_version) override; _FORCE_INLINE_ bool shadow_atlas_owns_light_instance(RID p_atlas, RID p_light_intance) { ShadowAtlas *atlas = shadow_atlas_owner.getornull(p_atlas); ERR_FAIL_COND_V(!atlas, false); @@ -772,9 +831,9 @@ public: return Size2(atlas->size, atlas->size); } - void directional_shadow_atlas_set_size(int p_size, bool p_16_bits = false); - int get_directional_light_shadow_size(RID p_light_intance); - void set_directional_shadow_count(int p_count); + virtual void directional_shadow_atlas_set_size(int p_size, bool p_16_bits = false) override; + virtual int get_directional_light_shadow_size(RID p_light_intance) override; + virtual void set_directional_shadow_count(int p_count) override; _FORCE_INLINE_ RID directional_shadow_get_texture() { return directional_shadow.depth; @@ -786,43 +845,43 @@ public: /* SDFGI UPDATE */ - virtual void sdfgi_update(RID p_render_buffers, RID p_environment, const Vector3 &p_world_position); - virtual int sdfgi_get_pending_region_count(RID p_render_buffers) const; - virtual AABB sdfgi_get_pending_region_bounds(RID p_render_buffers, int p_region) const; - virtual uint32_t sdfgi_get_pending_region_cascade(RID p_render_buffers, int p_region) const; + virtual void sdfgi_update(RID p_render_buffers, RID p_environment, const Vector3 &p_world_position) override; + virtual int sdfgi_get_pending_region_count(RID p_render_buffers) const override; + virtual AABB sdfgi_get_pending_region_bounds(RID p_render_buffers, int p_region) const override; + virtual uint32_t sdfgi_get_pending_region_cascade(RID p_render_buffers, int p_region) const override; RID sdfgi_get_ubo() const { return gi.sdfgi_ubo; } /* SKY API */ - virtual RID sky_allocate(); - virtual void sky_initialize(RID p_rid); + virtual RID sky_allocate() override; + virtual void sky_initialize(RID p_rid) override; - void sky_set_radiance_size(RID p_sky, int p_radiance_size); - void sky_set_mode(RID p_sky, RS::SkyMode p_mode); - void sky_set_material(RID p_sky, RID p_material); - Ref<Image> sky_bake_panorama(RID p_sky, float p_energy, bool p_bake_irradiance, const Size2i &p_size); + virtual void sky_set_radiance_size(RID p_sky, int p_radiance_size) override; + virtual void sky_set_mode(RID p_sky, RS::SkyMode p_mode) override; + virtual void sky_set_material(RID p_sky, RID p_material) override; + virtual Ref<Image> sky_bake_panorama(RID p_sky, float p_energy, bool p_bake_irradiance, const Size2i &p_size) override; /* ENVIRONMENT API */ - virtual RID environment_allocate(); - virtual void environment_initialize(RID p_rid); + virtual RID environment_allocate() override; + virtual void environment_initialize(RID p_rid) override; - void environment_set_background(RID p_env, RS::EnvironmentBG p_bg); - void environment_set_sky(RID p_env, RID p_sky); - void environment_set_sky_custom_fov(RID p_env, float p_scale); - void environment_set_sky_orientation(RID p_env, const Basis &p_orientation); - void environment_set_bg_color(RID p_env, const Color &p_color); - void environment_set_bg_energy(RID p_env, float p_energy); - void environment_set_canvas_max_layer(RID p_env, int p_max_layer); - void environment_set_ambient_light(RID p_env, const Color &p_color, RS::EnvironmentAmbientSource p_ambient = RS::ENV_AMBIENT_SOURCE_BG, float p_energy = 1.0, float p_sky_contribution = 0.0, RS::EnvironmentReflectionSource p_reflection_source = RS::ENV_REFLECTION_SOURCE_BG, const Color &p_ao_color = Color()); + virtual void environment_set_background(RID p_env, RS::EnvironmentBG p_bg) override; + virtual void environment_set_sky(RID p_env, RID p_sky) override; + virtual void environment_set_sky_custom_fov(RID p_env, float p_scale) override; + virtual void environment_set_sky_orientation(RID p_env, const Basis &p_orientation) override; + virtual void environment_set_bg_color(RID p_env, const Color &p_color) override; + virtual void environment_set_bg_energy(RID p_env, float p_energy) override; + virtual void environment_set_canvas_max_layer(RID p_env, int p_max_layer) override; + virtual void environment_set_ambient_light(RID p_env, const Color &p_color, RS::EnvironmentAmbientSource p_ambient = RS::ENV_AMBIENT_SOURCE_BG, float p_energy = 1.0, float p_sky_contribution = 0.0, RS::EnvironmentReflectionSource p_reflection_source = RS::ENV_REFLECTION_SOURCE_BG, const Color &p_ao_color = Color()) override; - RS::EnvironmentBG environment_get_background(RID p_env) const; + virtual RS::EnvironmentBG environment_get_background(RID p_env) const override; RID environment_get_sky(RID p_env) const; float environment_get_sky_custom_fov(RID p_env) const; Basis environment_get_sky_orientation(RID p_env) const; Color environment_get_bg_color(RID p_env) const; float environment_get_bg_energy(RID p_env) const; - int environment_get_canvas_max_layer(RID p_env) const; + virtual int environment_get_canvas_max_layer(RID p_env) const override; Color environment_get_ambient_light_color(RID p_env) const; RS::EnvironmentAmbientSource environment_get_ambient_source(RID p_env) const; float environment_get_ambient_light_energy(RID p_env) const; @@ -830,13 +889,13 @@ public: RS::EnvironmentReflectionSource environment_get_reflection_source(RID p_env) const; Color environment_get_ao_color(RID p_env) const; - bool is_environment(RID p_env) const; + virtual bool is_environment(RID p_env) const override; - void environment_set_glow(RID p_env, bool p_enable, Vector<float> p_levels, float p_intensity, float p_strength, float p_mix, float p_bloom_threshold, RS::EnvironmentGlowBlendMode p_blend_mode, float p_hdr_bleed_threshold, float p_hdr_bleed_scale, float p_hdr_luminance_cap); - void environment_glow_set_use_bicubic_upscale(bool p_enable); - void environment_glow_set_use_high_quality(bool p_enable); + virtual void environment_set_glow(RID p_env, bool p_enable, Vector<float> p_levels, float p_intensity, float p_strength, float p_mix, float p_bloom_threshold, RS::EnvironmentGlowBlendMode p_blend_mode, float p_hdr_bleed_threshold, float p_hdr_bleed_scale, float p_hdr_luminance_cap) override; + virtual void environment_glow_set_use_bicubic_upscale(bool p_enable) override; + virtual void environment_glow_set_use_high_quality(bool p_enable) override; - void environment_set_fog(RID p_env, bool p_enable, const Color &p_light_color, float p_light_energy, float p_sun_scatter, float p_density, float p_height, float p_height_density, float p_aerial_perspective); + virtual void environment_set_fog(RID p_env, bool p_enable, const Color &p_light_color, float p_light_energy, float p_sun_scatter, float p_density, float p_height, float p_height_density, float p_aerial_perspective) override; bool environment_is_fog_enabled(RID p_env) const; Color environment_get_fog_light_color(RID p_env) const; float environment_get_fog_light_energy(RID p_env) const; @@ -846,47 +905,53 @@ public: float environment_get_fog_height_density(RID p_env) const; float environment_get_fog_aerial_perspective(RID p_env) const; - void environment_set_volumetric_fog(RID p_env, bool p_enable, float p_density, const Color &p_light, float p_light_energy, float p_length, float p_detail_spread, float p_gi_inject, bool p_temporal_reprojection, float p_temporal_reprojection_amount); + virtual void environment_set_volumetric_fog(RID p_env, bool p_enable, float p_density, const Color &p_light, float p_light_energy, float p_length, float p_detail_spread, float p_gi_inject, bool p_temporal_reprojection, float p_temporal_reprojection_amount) override; - virtual void environment_set_volumetric_fog_volume_size(int p_size, int p_depth); - virtual void environment_set_volumetric_fog_filter_active(bool p_enable); + virtual void environment_set_volumetric_fog_volume_size(int p_size, int p_depth) override; + virtual void environment_set_volumetric_fog_filter_active(bool p_enable) override; - void environment_set_ssr(RID p_env, bool p_enable, int p_max_steps, float p_fade_int, float p_fade_out, float p_depth_tolerance); - void environment_set_ssao(RID p_env, bool p_enable, float p_radius, float p_intensity, float p_power, float p_detail, float p_horizon, float p_sharpness, float p_light_affect, float p_ao_channel_affect); - void environment_set_ssao_quality(RS::EnvironmentSSAOQuality p_quality, bool p_half_size, float p_adaptive_target, int p_blur_passes, float p_fadeout_from, float p_fadeout_to); + virtual void environment_set_ssr(RID p_env, bool p_enable, int p_max_steps, float p_fade_int, float p_fade_out, float p_depth_tolerance) override; + virtual void environment_set_ssao(RID p_env, bool p_enable, float p_radius, float p_intensity, float p_power, float p_detail, float p_horizon, float p_sharpness, float p_light_affect, float p_ao_channel_affect) override; + virtual void environment_set_ssao_quality(RS::EnvironmentSSAOQuality p_quality, bool p_half_size, float p_adaptive_target, int p_blur_passes, float p_fadeout_from, float p_fadeout_to) override; bool environment_is_ssao_enabled(RID p_env) const; float environment_get_ssao_ao_affect(RID p_env) const; float environment_get_ssao_light_affect(RID p_env) const; bool environment_is_ssr_enabled(RID p_env) const; bool environment_is_sdfgi_enabled(RID p_env) const; - virtual void environment_set_sdfgi(RID p_env, bool p_enable, RS::EnvironmentSDFGICascades p_cascades, float p_min_cell_size, RS::EnvironmentSDFGIYScale p_y_scale, bool p_use_occlusion, float p_bounce_feedback, bool p_read_sky, float p_energy, float p_normal_bias, float p_probe_bias); - virtual void environment_set_sdfgi_ray_count(RS::EnvironmentSDFGIRayCount p_ray_count); - virtual void environment_set_sdfgi_frames_to_converge(RS::EnvironmentSDFGIFramesToConverge p_frames); - virtual void environment_set_sdfgi_frames_to_update_light(RS::EnvironmentSDFGIFramesToUpdateLight p_update); + virtual void environment_set_sdfgi(RID p_env, bool p_enable, RS::EnvironmentSDFGICascades p_cascades, float p_min_cell_size, RS::EnvironmentSDFGIYScale p_y_scale, bool p_use_occlusion, float p_bounce_feedback, bool p_read_sky, float p_energy, float p_normal_bias, float p_probe_bias) override; + virtual void environment_set_sdfgi_ray_count(RS::EnvironmentSDFGIRayCount p_ray_count) override; + virtual void environment_set_sdfgi_frames_to_converge(RS::EnvironmentSDFGIFramesToConverge p_frames) override; + virtual void environment_set_sdfgi_frames_to_update_light(RS::EnvironmentSDFGIFramesToUpdateLight p_update) override; - void environment_set_ssr_roughness_quality(RS::EnvironmentSSRRoughnessQuality p_quality); + virtual void environment_set_ssr_roughness_quality(RS::EnvironmentSSRRoughnessQuality p_quality) override; RS::EnvironmentSSRRoughnessQuality environment_get_ssr_roughness_quality() const; - void environment_set_tonemap(RID p_env, RS::EnvironmentToneMapper p_tone_mapper, float p_exposure, float p_white, bool p_auto_exposure, float p_min_luminance, float p_max_luminance, float p_auto_exp_speed, float p_auto_exp_scale); - void environment_set_adjustment(RID p_env, bool p_enable, float p_brightness, float p_contrast, float p_saturation, bool p_use_1d_color_correction, RID p_color_correction); + virtual void environment_set_tonemap(RID p_env, RS::EnvironmentToneMapper p_tone_mapper, float p_exposure, float p_white, bool p_auto_exposure, float p_min_luminance, float p_max_luminance, float p_auto_exp_speed, float p_auto_exp_scale) override; + virtual void environment_set_adjustment(RID p_env, bool p_enable, float p_brightness, float p_contrast, float p_saturation, bool p_use_1d_color_correction, RID p_color_correction) override; + + virtual Ref<Image> environment_bake_panorama(RID p_env, bool p_bake_irradiance, const Size2i &p_size) override; + + virtual RID camera_effects_allocate() override; + virtual void camera_effects_initialize(RID p_rid) override; - virtual Ref<Image> environment_bake_panorama(RID p_env, bool p_bake_irradiance, const Size2i &p_size); + virtual void camera_effects_set_dof_blur_quality(RS::DOFBlurQuality p_quality, bool p_use_jitter) override; + virtual void camera_effects_set_dof_blur_bokeh_shape(RS::DOFBokehShape p_shape) override; - virtual RID camera_effects_allocate(); - virtual void camera_effects_initialize(RID p_rid); + virtual void camera_effects_set_dof_blur(RID p_camera_effects, bool p_far_enable, float p_far_distance, float p_far_transition, bool p_near_enable, float p_near_distance, float p_near_transition, float p_amount) override; + virtual void camera_effects_set_custom_exposure(RID p_camera_effects, bool p_enable, float p_exposure) override; - virtual void camera_effects_set_dof_blur_quality(RS::DOFBlurQuality p_quality, bool p_use_jitter); - virtual void camera_effects_set_dof_blur_bokeh_shape(RS::DOFBokehShape p_shape); + bool camera_effects_uses_dof(RID p_camera_effects) { + CameraEffects *camfx = camera_effects_owner.getornull(p_camera_effects); - virtual void camera_effects_set_dof_blur(RID p_camera_effects, bool p_far_enable, float p_far_distance, float p_far_transition, bool p_near_enable, float p_near_distance, float p_near_transition, float p_amount); - virtual void camera_effects_set_custom_exposure(RID p_camera_effects, bool p_enable, float p_exposure); + return camfx && (camfx->dof_blur_near_enabled || camfx->dof_blur_far_enabled) && camfx->dof_blur_amount > 0.0; + } - RID light_instance_create(RID p_light); - void light_instance_set_transform(RID p_light_instance, const Transform3D &p_transform); - void light_instance_set_aabb(RID p_light_instance, const AABB &p_aabb); - void light_instance_set_shadow_transform(RID p_light_instance, const CameraMatrix &p_projection, const Transform3D &p_transform, float p_far, float p_split, int p_pass, float p_shadow_texel_size, float p_bias_scale = 1.0, float p_range_begin = 0, const Vector2 &p_uv_scale = Vector2()); - void light_instance_mark_visible(RID p_light_instance); + virtual RID light_instance_create(RID p_light) override; + virtual void light_instance_set_transform(RID p_light_instance, const Transform3D &p_transform) override; + virtual void light_instance_set_aabb(RID p_light_instance, const AABB &p_aabb) override; + virtual void light_instance_set_shadow_transform(RID p_light_instance, const CameraMatrix &p_projection, const Transform3D &p_transform, float p_far, float p_split, int p_pass, float p_shadow_texel_size, float p_bias_scale = 1.0, float p_range_begin = 0, const Vector2 &p_uv_scale = Vector2()) override; + virtual void light_instance_mark_visible(RID p_light_instance) override; _FORCE_INLINE_ RID light_instance_get_base_light(RID p_light_instance) { LightInstance *li = light_instance_owner.getornull(p_light_instance); @@ -898,7 +963,7 @@ public: return li->transform; } - _FORCE_INLINE_ Rect2 light_instance_get_shadow_atlas_rect(RID p_light_instance, RID p_shadow_atlas) { + _FORCE_INLINE_ Rect2 light_instance_get_shadow_atlas_rect(RID p_light_instance, RID p_shadow_atlas, Vector2i &r_omni_offset) { ShadowAtlas *shadow_atlas = shadow_atlas_owner.getornull(p_shadow_atlas); LightInstance *li = light_instance_owner.getornull(p_light_instance); uint32_t key = shadow_atlas->shadow_owners[li->self]; @@ -918,6 +983,16 @@ public: x += (shadow % shadow_atlas->quadrants[quadrant].subdivision) * shadow_size; y += (shadow / shadow_atlas->quadrants[quadrant].subdivision) * shadow_size; + if (key & ShadowAtlas::OMNI_LIGHT_FLAG) { + if (((shadow + 1) % shadow_atlas->quadrants[quadrant].subdivision) == 0) { + r_omni_offset.x = 1 - int(shadow_atlas->quadrants[quadrant].subdivision); + r_omni_offset.y = 1; + } else { + r_omni_offset.x = 1; + r_omni_offset.y = 0; + } + } + uint32_t width = shadow_size; uint32_t height = shadow_size; @@ -998,14 +1073,9 @@ public: return li->last_pass; } - _FORCE_INLINE_ void light_instance_set_index(RID p_light_instance, uint32_t p_index) { + _FORCE_INLINE_ ForwardID light_instance_get_forward_id(RID p_light_instance) { LightInstance *li = light_instance_owner.getornull(p_light_instance); - li->light_index = p_index; - } - - _FORCE_INLINE_ uint32_t light_instance_get_index(RID p_light_instance) { - LightInstance *li = light_instance_owner.getornull(p_light_instance); - return li->light_index; + return li->forward_id; } _FORCE_INLINE_ RS::LightType light_instance_get_type(RID p_light_instance) { @@ -1013,9 +1083,9 @@ public: return li->light_type; } - virtual RID reflection_atlas_create(); - virtual void reflection_atlas_set_size(RID p_ref_atlas, int p_reflection_size, int p_reflection_count); - virtual int reflection_atlas_get_size(RID p_ref_atlas) const; + virtual RID reflection_atlas_create() override; + virtual void reflection_atlas_set_size(RID p_ref_atlas, int p_reflection_size, int p_reflection_count) override; + virtual int reflection_atlas_get_size(RID p_ref_atlas) const override; _FORCE_INLINE_ RID reflection_atlas_get_texture(RID p_ref_atlas) { ReflectionAtlas *atlas = reflection_atlas_owner.getornull(p_ref_atlas); @@ -1023,13 +1093,14 @@ public: return atlas->reflection; } - virtual RID reflection_probe_instance_create(RID p_probe); - virtual void reflection_probe_instance_set_transform(RID p_instance, const Transform3D &p_transform); - virtual void reflection_probe_release_atlas_index(RID p_instance); - virtual bool reflection_probe_instance_needs_redraw(RID p_instance); - virtual bool reflection_probe_instance_has_reflection(RID p_instance); - virtual bool reflection_probe_instance_begin_render(RID p_instance, RID p_reflection_atlas); - virtual bool reflection_probe_instance_postprocess_step(RID p_instance); + virtual RID reflection_probe_instance_create(RID p_probe) override; + virtual void reflection_probe_instance_set_transform(RID p_instance, const Transform3D &p_transform) override; + virtual void reflection_probe_release_atlas_index(RID p_instance) override; + virtual bool reflection_probe_instance_needs_redraw(RID p_instance) override; + virtual bool reflection_probe_instance_has_reflection(RID p_instance) override; + virtual bool reflection_probe_instance_begin_render(RID p_instance, RID p_reflection_atlas) override; + virtual RID reflection_probe_create_framebuffer(RID p_color, RID p_depth); + virtual bool reflection_probe_instance_postprocess_step(RID p_instance) override; uint32_t reflection_probe_instance_get_resolution(RID p_instance); RID reflection_probe_instance_get_framebuffer(RID p_instance, int p_index); @@ -1042,17 +1113,11 @@ public: return rpi->probe; } - _FORCE_INLINE_ void reflection_probe_instance_set_render_index(RID p_instance, uint32_t p_render_index) { - ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(p_instance); - ERR_FAIL_COND(!rpi); - rpi->render_index = p_render_index; - } - - _FORCE_INLINE_ uint32_t reflection_probe_instance_get_render_index(RID p_instance) { + _FORCE_INLINE_ ForwardID reflection_probe_instance_get_forward_id(RID p_instance) { ReflectionProbeInstance *rpi = reflection_probe_instance_owner.getornull(p_instance); ERR_FAIL_COND_V(!rpi, 0); - return rpi->render_index; + return rpi->forward_id; } _FORCE_INLINE_ void reflection_probe_instance_set_render_pass(RID p_instance, uint32_t p_render_pass) { @@ -1082,21 +1147,26 @@ public: return rpi->atlas_index; } - virtual RID decal_instance_create(RID p_decal); - virtual void decal_instance_set_transform(RID p_decal, const Transform3D &p_transform); + virtual RID decal_instance_create(RID p_decal) override; + virtual void decal_instance_set_transform(RID p_decal, const Transform3D &p_transform) override; _FORCE_INLINE_ RID decal_instance_get_base(RID p_decal) const { DecalInstance *decal = decal_instance_owner.getornull(p_decal); return decal->decal; } + _FORCE_INLINE_ ForwardID decal_instance_get_forward_id(RID p_decal) const { + DecalInstance *decal = decal_instance_owner.getornull(p_decal); + return decal->forward_id; + } + _FORCE_INLINE_ Transform3D decal_instance_get_transform(RID p_decal) const { DecalInstance *decal = decal_instance_owner.getornull(p_decal); return decal->transform; } - virtual RID lightmap_instance_create(RID p_lightmap); - virtual void lightmap_instance_set_transform(RID p_lightmap, const Transform3D &p_transform); + virtual RID lightmap_instance_create(RID p_lightmap) override; + virtual void lightmap_instance_set_transform(RID p_lightmap, const Transform3D &p_transform) override; _FORCE_INLINE_ bool lightmap_instance_is_valid(RID p_lightmap_instance) { return lightmap_instance_owner.getornull(p_lightmap_instance) != nullptr; } @@ -1110,24 +1180,26 @@ public: return li->transform; } - void _fill_instance_indices(const RID *p_omni_light_instances, uint32_t p_omni_light_instance_count, uint32_t *p_omni_light_indices, const RID *p_spot_light_instances, uint32_t p_spot_light_instance_count, uint32_t *p_spot_light_indices, const RID *p_reflection_probe_instances, uint32_t p_reflection_probe_instance_count, uint32_t *p_reflection_probe_indices, const RID *p_decal_instances, uint32_t p_decal_instance_count, uint32_t *p_decal_instance_indices, uint32_t p_layer_mask, uint32_t p_max_dst_words = 2); - /* gi light probes */ - RID voxel_gi_instance_create(RID p_base); - void voxel_gi_instance_set_transform_to_data(RID p_probe, const Transform3D &p_xform); - bool voxel_gi_needs_update(RID p_probe) const; - void voxel_gi_update(RID p_probe, bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<RendererSceneRender::GeometryInstance *> &p_dynamic_objects); - void voxel_gi_set_quality(RS::VoxelGIQuality p_quality) { gi.voxel_gi_quality = p_quality; } + virtual RID voxel_gi_instance_create(RID p_base) override; + virtual void voxel_gi_instance_set_transform_to_data(RID p_probe, const Transform3D &p_xform) override; + virtual bool voxel_gi_needs_update(RID p_probe) const override; + virtual void voxel_gi_update(RID p_probe, bool p_update_light_instances, const Vector<RID> &p_light_instances, const PagedArray<RendererSceneRender::GeometryInstance *> &p_dynamic_objects) override; + virtual void voxel_gi_set_quality(RS::VoxelGIQuality p_quality) override { gi.voxel_gi_quality = p_quality; } /* render buffers */ - RID render_buffers_create(); - void render_buffers_configure(RID p_render_buffers, RID p_render_target, int p_width, int p_height, RS::ViewportMSAA p_msaa, RS::ViewportScreenSpaceAA p_screen_space_aa, bool p_use_debanding, uint32_t p_view_count); - void gi_set_use_half_resolution(bool p_enable); + virtual float _render_buffers_get_luminance_multiplier(); + virtual RD::DataFormat _render_buffers_get_color_format(); + virtual bool _render_buffers_can_be_storage(); + virtual RID render_buffers_create() override; + virtual void render_buffers_configure(RID p_render_buffers, RID p_render_target, int p_width, int p_height, RS::ViewportMSAA p_msaa, RS::ViewportScreenSpaceAA p_screen_space_aa, bool p_use_debanding, uint32_t p_view_count) override; + virtual void gi_set_use_half_resolution(bool p_enable) override; RID render_buffers_get_ao_texture(RID p_render_buffers); RID render_buffers_get_back_buffer_texture(RID p_render_buffers); + RID render_buffers_get_back_depth_texture(RID p_render_buffers); RID render_buffers_get_voxel_gi_buffer(RID p_render_buffers); RID render_buffers_get_default_voxel_gi_buffer(); RID render_buffers_get_gi_ambient_texture(RID p_render_buffers); @@ -1152,30 +1224,34 @@ public: float render_buffers_get_volumetric_fog_end(RID p_render_buffers); float render_buffers_get_volumetric_fog_detail_spread(RID p_render_buffers); - void render_scene(RID p_render_buffers, const CameraData *p_camera_data, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_lights, const PagedArray<RID> &p_reflection_probes, const PagedArray<RID> &p_voxel_gi_instances, const PagedArray<RID> &p_decals, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_occluder_debug_tex, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold, const RenderShadowData *p_render_shadows, int p_render_shadow_count, const RenderSDFGIData *p_render_sdfgi_regions, int p_render_sdfgi_region_count, const RenderSDFGIUpdateData *p_sdfgi_update_data = nullptr); + virtual void render_scene(RID p_render_buffers, const CameraData *p_camera_data, const PagedArray<GeometryInstance *> &p_instances, const PagedArray<RID> &p_lights, const PagedArray<RID> &p_reflection_probes, const PagedArray<RID> &p_voxel_gi_instances, const PagedArray<RID> &p_decals, const PagedArray<RID> &p_lightmaps, RID p_environment, RID p_camera_effects, RID p_shadow_atlas, RID p_occluder_debug_tex, RID p_reflection_atlas, RID p_reflection_probe, int p_reflection_probe_pass, float p_screen_lod_threshold, const RenderShadowData *p_render_shadows, int p_render_shadow_count, const RenderSDFGIData *p_render_sdfgi_regions, int p_render_sdfgi_region_count, const RenderSDFGIUpdateData *p_sdfgi_update_data = nullptr, RendererScene::RenderInfo *r_render_info = nullptr) override; - void render_material(const Transform3D &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region); + virtual void render_material(const Transform3D &p_cam_transform, const CameraMatrix &p_cam_projection, bool p_cam_ortogonal, const PagedArray<GeometryInstance *> &p_instances, RID p_framebuffer, const Rect2i &p_region) override; - void render_particle_collider_heightfield(RID p_collider, const Transform3D &p_transform, const PagedArray<GeometryInstance *> &p_instances); + virtual void render_particle_collider_heightfield(RID p_collider, const Transform3D &p_transform, const PagedArray<GeometryInstance *> &p_instances) override; - virtual void set_scene_pass(uint64_t p_pass) { + virtual void set_scene_pass(uint64_t p_pass) override { scene_pass = p_pass; } _FORCE_INLINE_ uint64_t get_scene_pass() { return scene_pass; } - virtual void screen_space_roughness_limiter_set_active(bool p_enable, float p_amount, float p_limit); - virtual bool screen_space_roughness_limiter_is_active() const; + virtual void screen_space_roughness_limiter_set_active(bool p_enable, float p_amount, float p_limit) override; + virtual bool screen_space_roughness_limiter_is_active() const override; virtual float screen_space_roughness_limiter_get_amount() const; virtual float screen_space_roughness_limiter_get_limit() const; - virtual void sub_surface_scattering_set_quality(RS::SubSurfaceScatteringQuality p_quality); + virtual void sub_surface_scattering_set_quality(RS::SubSurfaceScatteringQuality p_quality) override; RS::SubSurfaceScatteringQuality sub_surface_scattering_get_quality() const; - virtual void sub_surface_scattering_set_scale(float p_scale, float p_depth_scale); + virtual void sub_surface_scattering_set_scale(float p_scale, float p_depth_scale) override; + + virtual void shadows_quality_set(RS::ShadowQuality p_quality) override; + virtual void directional_shadow_quality_set(RS::ShadowQuality p_quality) override; + + virtual void decals_set_filter(RS::DecalFilter p_filter) override; + virtual void light_projectors_set_filter(RS::LightProjectorFilter p_filter) override; - virtual void shadows_quality_set(RS::ShadowQuality p_quality); - virtual void directional_shadow_quality_set(RS::ShadowQuality p_quality); _FORCE_INLINE_ RS::ShadowQuality shadows_quality_get() const { return shadows_quality; } _FORCE_INLINE_ RS::ShadowQuality directional_shadow_quality_get() const { return directional_shadow_quality; } _FORCE_INLINE_ float shadows_quality_radius_get() const { return shadows_quality_radius; } @@ -1191,21 +1267,24 @@ public: _FORCE_INLINE_ int penumbra_shadow_samples_get() const { return penumbra_shadow_samples; } _FORCE_INLINE_ int soft_shadow_samples_get() const { return soft_shadow_samples; } + _FORCE_INLINE_ RS::LightProjectorFilter light_projectors_get_filter() const { return light_projectors_filter; } + _FORCE_INLINE_ RS::DecalFilter decals_get_filter() const { return decals_filter; } + int get_roughness_layers() const; bool is_using_radiance_cubemap_array() const; - virtual TypedArray<Image> bake_render_uv2(RID p_base, const Vector<RID> &p_material_overrides, const Size2i &p_image_size); + virtual TypedArray<Image> bake_render_uv2(RID p_base, const Vector<RID> &p_material_overrides, const Size2i &p_image_size) override; - virtual bool free(RID p_rid); + virtual bool free(RID p_rid) override; - virtual void update(); + virtual void update() override; - virtual void set_debug_draw_mode(RS::ViewportDebugDraw p_debug_draw); + virtual void set_debug_draw_mode(RS::ViewportDebugDraw p_debug_draw) override; _FORCE_INLINE_ RS::ViewportDebugDraw get_debug_draw_mode() const { return debug_draw; } - void set_time(double p_time, double p_step); + virtual void set_time(double p_time, double p_step) override; RID get_reflection_probe_buffer(); RID get_omni_light_buffer(); @@ -1214,13 +1293,15 @@ public: RID get_decal_buffer(); int get_max_directional_lights() const; - void sdfgi_set_debug_probe_select(const Vector3 &p_position, const Vector3 &p_dir); + virtual void sdfgi_set_debug_probe_select(const Vector3 &p_position, const Vector3 &p_dir) override; virtual bool is_dynamic_gi_supported() const; virtual bool is_clustered_enabled() const; virtual bool is_volumetric_supported() const; virtual uint32_t get_max_elements() const; + void init(); + RendererSceneRenderRD(RendererStorageRD *p_storage); ~RendererSceneRenderRD(); }; diff --git a/servers/rendering/renderer_rd/renderer_scene_sky_rd.cpp b/servers/rendering/renderer_rd/renderer_scene_sky_rd.cpp index 1aa01dd16e..c388da755c 100644 --- a/servers/rendering/renderer_rd/renderer_scene_sky_rd.cpp +++ b/servers/rendering/renderer_rd/renderer_scene_sky_rd.cpp @@ -92,8 +92,7 @@ void RendererSceneSkyRD::SkyShaderData::set_code(const String &p_code) { RendererSceneRenderRD *scene_singleton = (RendererSceneRenderRD *)RendererSceneRenderRD::singleton; Error err = scene_singleton->sky.sky_shader.compiler.compile(RS::SHADER_SKY, code, &actions, path, gen_code); - - ERR_FAIL_COND(err != OK); + ERR_FAIL_COND_MSG(err != OK, "Shader compilation failed."); if (version.is_null()) { version = scene_singleton->sky.sky_shader.shader.version_create(); @@ -230,96 +229,77 @@ RendererSceneSkyRD::SkyShaderData::~SkyShaderData() { //////////////////////////////////////////////////////////////////////////////// // Sky material -void RendererSceneSkyRD::SkyMaterialData::update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) { +bool RendererSceneSkyRD::SkyMaterialData::update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) { RendererSceneRenderRD *scene_singleton = (RendererSceneRenderRD *)RendererSceneRenderRD::singleton; uniform_set_updated = true; - if ((uint32_t)ubo_data.size() != shader_data->ubo_size) { - p_uniform_dirty = true; - if (uniform_buffer.is_valid()) { - RD::get_singleton()->free(uniform_buffer); - uniform_buffer = RID(); - } - - ubo_data.resize(shader_data->ubo_size); - if (ubo_data.size()) { - uniform_buffer = RD::get_singleton()->uniform_buffer_create(ubo_data.size()); - memset(ubo_data.ptrw(), 0, ubo_data.size()); //clear - } - - //clear previous uniform set - if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - RD::get_singleton()->free(uniform_set); - uniform_set = RID(); - } - } + return update_parameters_uniform_set(p_parameters, p_uniform_dirty, p_textures_dirty, shader_data->uniforms, shader_data->ubo_offsets.ptr(), shader_data->texture_uniforms, shader_data->default_texture_params, shader_data->ubo_size, uniform_set, scene_singleton->sky.sky_shader.shader.version_get_shader(shader_data->version, 0), SKY_SET_MATERIAL); +} - //check whether buffer changed - if (p_uniform_dirty && ubo_data.size()) { - update_uniform_buffer(shader_data->uniforms, shader_data->ubo_offsets.ptr(), p_parameters, ubo_data.ptrw(), ubo_data.size(), false); - RD::get_singleton()->buffer_update(uniform_buffer, 0, ubo_data.size(), ubo_data.ptrw()); - } +RendererSceneSkyRD::SkyMaterialData::~SkyMaterialData() { + free_parameters_uniform_set(uniform_set); +} - uint32_t tex_uniform_count = shader_data->texture_uniforms.size(); +//////////////////////////////////////////////////////////////////////////////// +// Render sky + +static _FORCE_INLINE_ void store_transform_3x3(const Basis &p_basis, float *p_array) { + p_array[0] = p_basis.elements[0][0]; + p_array[1] = p_basis.elements[1][0]; + p_array[2] = p_basis.elements[2][0]; + p_array[3] = 0; + p_array[4] = p_basis.elements[0][1]; + p_array[5] = p_basis.elements[1][1]; + p_array[6] = p_basis.elements[2][1]; + p_array[7] = 0; + p_array[8] = p_basis.elements[0][2]; + p_array[9] = p_basis.elements[1][2]; + p_array[10] = p_basis.elements[2][2]; + p_array[11] = 0; +} - if ((uint32_t)texture_cache.size() != tex_uniform_count) { - texture_cache.resize(tex_uniform_count); - p_textures_dirty = true; +void RendererSceneSkyRD::_render_sky(RD::DrawListID p_list, float p_time, RID p_fb, PipelineCacheRD *p_pipeline, RID p_uniform_set, RID p_texture_set, uint32_t p_view_count, const CameraMatrix *p_projections, const Basis &p_orientation, float p_multiplier, const Vector3 &p_position, float p_luminance_multiplier) { + SkyPushConstant sky_push_constant; - //clear previous uniform set - if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - RD::get_singleton()->free(uniform_set); - uniform_set = RID(); - } - } + memset(&sky_push_constant, 0, sizeof(SkyPushConstant)); - if (p_textures_dirty && tex_uniform_count) { - update_textures(p_parameters, shader_data->default_texture_params, shader_data->texture_uniforms, texture_cache.ptrw(), true); + for (uint32_t v = 0; v < p_view_count; v++) { + // We only need key components of our projection matrix + sky_push_constant.projections[v][0] = p_projections[v].matrix[2][0]; + sky_push_constant.projections[v][1] = p_projections[v].matrix[0][0]; + sky_push_constant.projections[v][2] = p_projections[v].matrix[2][1]; + sky_push_constant.projections[v][3] = p_projections[v].matrix[1][1]; } + sky_push_constant.position[0] = p_position.x; + sky_push_constant.position[1] = p_position.y; + sky_push_constant.position[2] = p_position.z; + sky_push_constant.multiplier = p_multiplier; + sky_push_constant.time = p_time; + sky_push_constant.luminance_multiplier = p_luminance_multiplier; + store_transform_3x3(p_orientation, sky_push_constant.orientation); - if (shader_data->ubo_size == 0 && shader_data->texture_uniforms.size() == 0) { - // This material does not require an uniform set, so don't create it. - return; - } + RenderingDevice::FramebufferFormatID fb_format = RD::get_singleton()->framebuffer_get_format(p_fb); - if (!p_textures_dirty && uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - //no reason to update uniform set, only UBO (or nothing) was needed to update - return; - } + RD::DrawListID draw_list = p_list; - Vector<RD::Uniform> uniforms; + RD::get_singleton()->draw_list_bind_render_pipeline(draw_list, p_pipeline->get_render_pipeline(RD::INVALID_ID, fb_format, false, RD::get_singleton()->draw_list_get_current_pass())); + // Update uniform sets. { - if (shader_data->ubo_size) { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; - u.binding = 0; - u.ids.push_back(uniform_buffer); - uniforms.push_back(u); - } - - const RID *textures = texture_cache.ptrw(); - for (uint32_t i = 0; i < tex_uniform_count; i++) { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.binding = 1 + i; - u.ids.push_back(textures[i]); - uniforms.push_back(u); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, sky_scene_state.uniform_set, 0); + if (p_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(p_uniform_set)) { // Material may not have a uniform set. + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, p_uniform_set, 1); } + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, p_texture_set, 2); + RD::get_singleton()->draw_list_bind_uniform_set(draw_list, sky_scene_state.fog_uniform_set, 3); } - uniform_set = RD::get_singleton()->uniform_set_create(uniforms, scene_singleton->sky.sky_shader.shader.version_get_shader(shader_data->version, 0), SKY_SET_MATERIAL); -} + RD::get_singleton()->draw_list_bind_index_array(draw_list, index_array); -RendererSceneSkyRD::SkyMaterialData::~SkyMaterialData() { - if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - RD::get_singleton()->free(uniform_set); - } + RD::get_singleton()->draw_list_set_push_constant(draw_list, &sky_push_constant, sizeof(SkyPushConstant)); - if (uniform_buffer.is_valid()) { - RD::get_singleton()->free(uniform_buffer); - } + RD::get_singleton()->draw_list_draw(draw_list, true); } //////////////////////////////////////////////////////////////////////////////// @@ -336,12 +316,16 @@ void RendererSceneSkyRD::ReflectionData::clear_reflection_data() { coefficient_buffer = RID(); } -void RendererSceneSkyRD::ReflectionData::update_reflection_data(int p_size, int p_mipmaps, bool p_use_array, RID p_base_cube, int p_base_layer, bool p_low_quality, int p_roughness_layers) { +void RendererSceneSkyRD::ReflectionData::update_reflection_data(RendererStorageRD *p_storage, int p_size, int p_mipmaps, bool p_use_array, RID p_base_cube, int p_base_layer, bool p_low_quality, int p_roughness_layers, RD::DataFormat p_texture_format) { //recreate radiance and all data int mipmaps = p_mipmaps; uint32_t w = p_size, h = p_size; + EffectsRD *effects = p_storage->get_effects(); + ERR_FAIL_NULL_MSG(effects, "Effects haven't been initialised"); + bool prefer_raster_effects = effects->get_prefer_raster_effects(); + if (p_use_array) { int num_layers = p_low_quality ? 8 : p_roughness_layers; @@ -400,9 +384,9 @@ void RendererSceneSkyRD::ReflectionData::update_reflection_data(int p_size, int } radiance_base_cubemap = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), p_base_cube, p_base_layer, 0, RD::TEXTURE_SLICE_CUBEMAP); - + RD::get_singleton()->set_resource_name(radiance_base_cubemap, "radiance base cubemap"); RD::TextureFormat tf; - tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + tf.format = p_texture_format; tf.width = 64; // Always 64x64 tf.height = 64; tf.texture_type = RD::TEXTURE_TYPE_CUBE; @@ -411,6 +395,7 @@ void RendererSceneSkyRD::ReflectionData::update_reflection_data(int p_size, int tf.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_STORAGE_BIT | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; downsampled_radiance_cubemap = RD::get_singleton()->texture_create(tf, RD::TextureView()); + RD::get_singleton()->set_resource_name(downsampled_radiance_cubemap, "downsampled radiance cubemap"); { uint32_t mmw = 64; uint32_t mmh = 64; @@ -420,6 +405,18 @@ void RendererSceneSkyRD::ReflectionData::update_reflection_data(int p_size, int mm.size.width = mmw; mm.size.height = mmh; mm.view = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), downsampled_radiance_cubemap, 0, j, RD::TEXTURE_SLICE_CUBEMAP); + RD::get_singleton()->set_resource_name(mm.view, "Downsampled Radiance Cubemap Mip " + itos(j) + " "); + if (prefer_raster_effects) { + // we need a framebuffer for each side of our cubemap + + for (int k = 0; k < 6; k++) { + mm.views[k] = RD::get_singleton()->texture_create_shared_from_slice(RD::TextureView(), downsampled_radiance_cubemap, k, j); + RD::get_singleton()->set_resource_name(mm.view, "Downsampled Radiance Cubemap Mip: " + itos(j) + " Face: " + itos(k) + " "); + Vector<RID> fbtex; + fbtex.push_back(mm.views[k]); + mm.framebuffers[k] = RD::get_singleton()->framebuffer_create(fbtex); + } + } mmw = MAX(1, mmw >> 1); mmh = MAX(1, mmh >> 1); @@ -428,50 +425,128 @@ void RendererSceneSkyRD::ReflectionData::update_reflection_data(int p_size, int } void RendererSceneSkyRD::ReflectionData::create_reflection_fast_filter(RendererStorageRD *p_storage, bool p_use_arrays) { - p_storage->get_effects()->cubemap_downsample(radiance_base_cubemap, downsampled_layer.mipmaps[0].view, downsampled_layer.mipmaps[0].size); + EffectsRD *effects = p_storage->get_effects(); + ERR_FAIL_NULL_MSG(effects, "Effects haven't been initialised"); + bool prefer_raster_effects = effects->get_prefer_raster_effects(); + + if (prefer_raster_effects) { + RD::get_singleton()->draw_command_begin_label("Downsample radiance map"); + for (int k = 0; k < 6; k++) { + effects->cubemap_downsample_raster(radiance_base_cubemap, downsampled_layer.mipmaps[0].framebuffers[k], k, downsampled_layer.mipmaps[0].size); + } - for (int i = 1; i < downsampled_layer.mipmaps.size(); i++) { - p_storage->get_effects()->cubemap_downsample(downsampled_layer.mipmaps[i - 1].view, downsampled_layer.mipmaps[i].view, downsampled_layer.mipmaps[i].size); - } + for (int i = 1; i < downsampled_layer.mipmaps.size(); i++) { + for (int k = 0; k < 6; k++) { + effects->cubemap_downsample_raster(downsampled_layer.mipmaps[i - 1].view, downsampled_layer.mipmaps[i].framebuffers[k], k, downsampled_layer.mipmaps[i].size); + } + } + RD::get_singleton()->draw_command_end_label(); // Downsample Radiance - Vector<RID> views; - if (p_use_arrays) { - for (int i = 1; i < layers.size(); i++) { - views.push_back(layers[i].views[0]); + if (p_use_arrays) { + RD::get_singleton()->draw_command_begin_label("filter radiance map into array heads"); + for (int i = 0; i < layers.size(); i++) { + for (int k = 0; k < 6; k++) { + effects->cubemap_filter_raster(downsampled_radiance_cubemap, layers[i].mipmaps[0].framebuffers[k], k, i); + } + } + } else { + RD::get_singleton()->draw_command_begin_label("filter radiance map into mipmaps directly"); + for (int j = 0; j < layers[0].mipmaps.size(); j++) { + for (int k = 0; k < 6; k++) { + effects->cubemap_filter_raster(downsampled_radiance_cubemap, layers[0].mipmaps[j].framebuffers[k], k, j); + } + } } + RD::get_singleton()->draw_command_end_label(); // Filter radiance } else { - for (int i = 1; i < layers[0].views.size(); i++) { - views.push_back(layers[0].views[i]); + effects->cubemap_downsample(radiance_base_cubemap, downsampled_layer.mipmaps[0].view, downsampled_layer.mipmaps[0].size); + + for (int i = 1; i < downsampled_layer.mipmaps.size(); i++) { + effects->cubemap_downsample(downsampled_layer.mipmaps[i - 1].view, downsampled_layer.mipmaps[i].view, downsampled_layer.mipmaps[i].size); + } + + Vector<RID> views; + if (p_use_arrays) { + for (int i = 1; i < layers.size(); i++) { + views.push_back(layers[i].views[0]); + } + } else { + for (int i = 1; i < layers[0].views.size(); i++) { + views.push_back(layers[0].views[i]); + } } - } - p_storage->get_effects()->cubemap_filter(downsampled_radiance_cubemap, views, p_use_arrays); + effects->cubemap_filter(downsampled_radiance_cubemap, views, p_use_arrays); + } } void RendererSceneSkyRD::ReflectionData::create_reflection_importance_sample(RendererStorageRD *p_storage, bool p_use_arrays, int p_cube_side, int p_base_layer, uint32_t p_sky_ggx_samples_quality) { - if (p_use_arrays) { - //render directly to the layers - p_storage->get_effects()->cubemap_roughness(radiance_base_cubemap, layers[p_base_layer].views[0], p_cube_side, p_sky_ggx_samples_quality, float(p_base_layer) / (layers.size() - 1.0), layers[p_base_layer].mipmaps[0].size.x); + EffectsRD *effects = p_storage->get_effects(); + ERR_FAIL_NULL_MSG(effects, "Effects haven't been initialised"); + bool prefer_raster_effects = effects->get_prefer_raster_effects(); + + if (prefer_raster_effects) { + // Need to ask clayjohn but p_cube_side is set to 10, looks like in the compute shader we're doing all 6 sides in one call + // here we need to do them one by one so ignoring p_cube_side + if (p_use_arrays) { + for (int k = 0; k < 6; k++) { + effects->cubemap_roughness_raster( + radiance_base_cubemap, + layers[p_base_layer].mipmaps[0].framebuffers[k], + k, + p_sky_ggx_samples_quality, + float(p_base_layer) / (layers.size() - 1.0), + layers[p_base_layer].mipmaps[0].size.x); + } + } else { + for (int k = 0; k < 6; k++) { + effects->cubemap_roughness_raster( + layers[0].views[p_base_layer - 1], + layers[0].mipmaps[p_base_layer].framebuffers[k], + k, + p_sky_ggx_samples_quality, + float(p_base_layer) / (layers[0].mipmaps.size() - 1.0), + layers[0].mipmaps[p_base_layer].size.x); + } + } } else { - p_storage->get_effects()->cubemap_roughness( - layers[0].views[p_base_layer - 1], - layers[0].views[p_base_layer], - p_cube_side, - p_sky_ggx_samples_quality, - float(p_base_layer) / (layers[0].mipmaps.size() - 1.0), - layers[0].mipmaps[p_base_layer].size.x); + if (p_use_arrays) { + //render directly to the layers + effects->cubemap_roughness(radiance_base_cubemap, layers[p_base_layer].views[0], p_cube_side, p_sky_ggx_samples_quality, float(p_base_layer) / (layers.size() - 1.0), layers[p_base_layer].mipmaps[0].size.x); + } else { + effects->cubemap_roughness( + layers[0].views[p_base_layer - 1], + layers[0].views[p_base_layer], + p_cube_side, + p_sky_ggx_samples_quality, + float(p_base_layer) / (layers[0].mipmaps.size() - 1.0), + layers[0].mipmaps[p_base_layer].size.x); + } } } void RendererSceneSkyRD::ReflectionData::update_reflection_mipmaps(RendererStorageRD *p_storage, int p_start, int p_end) { + EffectsRD *effects = p_storage->get_effects(); + ERR_FAIL_NULL_MSG(effects, "Effects haven't been initialised"); + bool prefer_raster_effects = effects->get_prefer_raster_effects(); + + RD::get_singleton()->draw_command_begin_label("Update Radiance Cubemap Array Mipmaps"); for (int i = p_start; i < p_end; i++) { for (int j = 0; j < layers[i].views.size() - 1; j++) { RID view = layers[i].views[j]; - RID texture = layers[i].views[j + 1]; Size2i size = layers[i].mipmaps[j + 1].size; - p_storage->get_effects()->cubemap_downsample(view, texture, size); + if (prefer_raster_effects) { + for (int k = 0; k < 6; k++) { + RID framebuffer = layers[i].mipmaps[j + 1].framebuffers[k]; + effects->cubemap_downsample_raster(view, framebuffer, k, size); + } + } else { + RID texture = layers[i].views[j + 1]; + effects->cubemap_downsample(view, texture, size); + } } } + RD::get_singleton()->draw_command_end_label(); } //////////////////////////////////////////////////////////////////////////////// @@ -628,7 +703,7 @@ Ref<Image> RendererSceneSkyRD::Sky::bake_panorama(RendererStorageRD *p_storage, RD::get_singleton()->free(rad_tex); Ref<Image> img; - img.instance(); + img.instantiate(); img->create(p_size.width, p_size.height, false, Image::FORMAT_RGBAF, data); for (int i = 0; i < p_size.width; i++) { for (int j = 0; j < p_size.height; j++) { @@ -780,7 +855,15 @@ void RendererSceneSkyRD::init(RendererStorageRD *p_storage) { sky_shader.default_shader = storage->shader_allocate(); storage->shader_initialize(sky_shader.default_shader); - storage->shader_set_code(sky_shader.default_shader, "shader_type sky; void sky() { COLOR = vec3(0.0); } \n"); + storage->shader_set_code(sky_shader.default_shader, R"( +// Default sky shader. + +shader_type sky; + +void sky() { + COLOR = vec3(0.0); +} +)"); sky_shader.default_material = storage->material_allocate(); storage->material_initialize(sky_shader.default_material); @@ -861,7 +944,17 @@ void RendererSceneSkyRD::init(RendererStorageRD *p_storage) { sky_scene_state.fog_shader = storage->shader_allocate(); storage->shader_initialize(sky_scene_state.fog_shader); - storage->shader_set_code(sky_scene_state.fog_shader, "shader_type sky; uniform vec4 clear_color; void sky() { COLOR = clear_color.rgb; } \n"); + storage->shader_set_code(sky_scene_state.fog_shader, R"( +// Default clear color sky shader. + +shader_type sky; + +uniform vec4 clear_color; + +void sky() { + COLOR = clear_color.rgb; +} +)"); sky_scene_state.fog_material = storage->material_allocate(); storage->material_initialize(sky_scene_state.fog_material); @@ -892,6 +985,45 @@ void RendererSceneSkyRD::init(RendererStorageRD *p_storage) { sky_scene_state.fog_only_texture_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, sky_shader.default_shader_rd, SKY_SET_TEXTURES); } + + { //create index array for copy shaders + Vector<uint8_t> pv; + pv.resize(6 * 4); + { + uint8_t *w = pv.ptrw(); + int *p32 = (int *)w; + p32[0] = 0; + p32[1] = 1; + p32[2] = 2; + p32[3] = 0; + p32[4] = 2; + p32[5] = 3; + } + index_buffer = RD::get_singleton()->index_buffer_create(6, RenderingDevice::INDEX_BUFFER_FORMAT_UINT32, pv); + index_array = RD::get_singleton()->index_array_create(index_buffer, 0, 6); + } +} + +void RendererSceneSkyRD::set_texture_format(RD::DataFormat p_texture_format) { + texture_format = p_texture_format; +} + +RendererSceneSkyRD::~RendererSceneSkyRD() { + // TODO cleanup anything created in init... + + if (RD::get_singleton()->uniform_set_is_valid(sky_scene_state.uniform_set)) { + RD::get_singleton()->free(sky_scene_state.uniform_set); + } + + if (RD::get_singleton()->uniform_set_is_valid(sky_scene_state.default_fog_uniform_set)) { + RD::get_singleton()->free(sky_scene_state.default_fog_uniform_set); + } + + if (RD::get_singleton()->uniform_set_is_valid(sky_scene_state.fog_only_texture_uniform_set)) { + RD::get_singleton()->free(sky_scene_state.fog_only_texture_uniform_set); + } + + RD::get_singleton()->free(index_buffer); //array gets freed as dependency } void RendererSceneSkyRD::setup(RendererSceneEnvironmentRD *p_env, RID p_render_buffers, const CameraMatrix &p_projection, const Transform3D &p_transform, const Size2i p_screen_size, RendererSceneRenderRD *p_scene_render) { @@ -1064,7 +1196,7 @@ void RendererSceneSkyRD::setup(RendererSceneEnvironmentRD *p_env, RID p_render_b RD::get_singleton()->buffer_update(sky_scene_state.uniform_buffer, 0, sizeof(SkySceneState::UBO), &sky_scene_state.ubo); } -void RendererSceneSkyRD::update(RendererSceneEnvironmentRD *p_env, const CameraMatrix &p_projection, const Transform3D &p_transform, double p_time) { +void RendererSceneSkyRD::update(RendererSceneEnvironmentRD *p_env, const CameraMatrix &p_projection, const Transform3D &p_transform, double p_time, float p_luminance_multiplier) { ERR_FAIL_COND(!p_env); Sky *sky = get_sky(p_env->sky); @@ -1144,6 +1276,7 @@ void RendererSceneSkyRD::update(RendererSceneEnvironmentRD *p_env, const CameraM cm = correction * cm; if (shader_data->uses_quarter_res) { + RD::get_singleton()->draw_command_begin_label("Render Sky to Quarter Res Cubemap"); PipelineCacheRD *pipeline = &shader_data->pipelines[SKY_VERSION_CUBEMAP_QUARTER_RES]; Vector<Color> clear_colors; @@ -1151,17 +1284,18 @@ void RendererSceneSkyRD::update(RendererSceneEnvironmentRD *p_env, const CameraM RD::DrawListID cubemap_draw_list; for (int i = 0; i < 6; i++) { - Transform3D local_view; - local_view.set_look_at(Vector3(0, 0, 0), view_normals[i], view_up[i]); + Basis local_view = Basis::looking_at(view_normals[i], view_up[i]); RID texture_uniform_set = sky->get_textures(storage, SKY_TEXTURE_SET_CUBEMAP_QUARTER_RES, sky_shader.default_shader_rd); cubemap_draw_list = RD::get_singleton()->draw_list_begin(sky->reflection.layers[0].mipmaps[2].framebuffers[i], RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); - storage->get_effects()->render_sky(cubemap_draw_list, p_time, sky->reflection.layers[0].mipmaps[2].framebuffers[i], sky_scene_state.uniform_set, sky_scene_state.fog_uniform_set, pipeline, material->uniform_set, texture_uniform_set, 1, &cm, local_view.basis, multiplier, p_transform.origin); + _render_sky(cubemap_draw_list, p_time, sky->reflection.layers[0].mipmaps[2].framebuffers[i], pipeline, material->uniform_set, texture_uniform_set, 1, &cm, local_view, multiplier, p_transform.origin, p_luminance_multiplier); RD::get_singleton()->draw_list_end(); } + RD::get_singleton()->draw_command_end_label(); } if (shader_data->uses_half_res) { + RD::get_singleton()->draw_command_begin_label("Render Sky to Half Res Cubemap"); PipelineCacheRD *pipeline = &shader_data->pipelines[SKY_VERSION_CUBEMAP_HALF_RES]; Vector<Color> clear_colors; @@ -1169,28 +1303,29 @@ void RendererSceneSkyRD::update(RendererSceneEnvironmentRD *p_env, const CameraM RD::DrawListID cubemap_draw_list; for (int i = 0; i < 6; i++) { - Transform3D local_view; - local_view.set_look_at(Vector3(0, 0, 0), view_normals[i], view_up[i]); + Basis local_view = Basis::looking_at(view_normals[i], view_up[i]); RID texture_uniform_set = sky->get_textures(storage, SKY_TEXTURE_SET_CUBEMAP_HALF_RES, sky_shader.default_shader_rd); cubemap_draw_list = RD::get_singleton()->draw_list_begin(sky->reflection.layers[0].mipmaps[1].framebuffers[i], RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); - storage->get_effects()->render_sky(cubemap_draw_list, p_time, sky->reflection.layers[0].mipmaps[1].framebuffers[i], sky_scene_state.uniform_set, sky_scene_state.fog_uniform_set, pipeline, material->uniform_set, texture_uniform_set, 1, &cm, local_view.basis, multiplier, p_transform.origin); + _render_sky(cubemap_draw_list, p_time, sky->reflection.layers[0].mipmaps[1].framebuffers[i], pipeline, material->uniform_set, texture_uniform_set, 1, &cm, local_view, multiplier, p_transform.origin, p_luminance_multiplier); RD::get_singleton()->draw_list_end(); } + RD::get_singleton()->draw_command_end_label(); } RD::DrawListID cubemap_draw_list; PipelineCacheRD *pipeline = &shader_data->pipelines[SKY_VERSION_CUBEMAP]; + RD::get_singleton()->draw_command_begin_label("Render Sky Cubemap"); for (int i = 0; i < 6; i++) { - Transform3D local_view; - local_view.set_look_at(Vector3(0, 0, 0), view_normals[i], view_up[i]); + Basis local_view = Basis::looking_at(view_normals[i], view_up[i]); RID texture_uniform_set = sky->get_textures(storage, SKY_TEXTURE_SET_CUBEMAP, sky_shader.default_shader_rd); cubemap_draw_list = RD::get_singleton()->draw_list_begin(sky->reflection.layers[0].mipmaps[0].framebuffers[i], RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_KEEP, RD::FINAL_ACTION_DISCARD); - storage->get_effects()->render_sky(cubemap_draw_list, p_time, sky->reflection.layers[0].mipmaps[0].framebuffers[i], sky_scene_state.uniform_set, sky_scene_state.fog_uniform_set, pipeline, material->uniform_set, texture_uniform_set, 1, &cm, local_view.basis, multiplier, p_transform.origin); + _render_sky(cubemap_draw_list, p_time, sky->reflection.layers[0].mipmaps[0].framebuffers[i], pipeline, material->uniform_set, texture_uniform_set, 1, &cm, local_view, multiplier, p_transform.origin, p_luminance_multiplier); RD::get_singleton()->draw_list_end(); } + RD::get_singleton()->draw_command_end_label(); if (sky_mode == RS::SKY_MODE_REALTIME) { sky->reflection.create_reflection_fast_filter(storage, sky_use_cubemap_array); @@ -1305,7 +1440,7 @@ void RendererSceneSkyRD::draw(RendererSceneEnvironmentRD *p_env, bool p_can_cont clear_colors.push_back(Color(0.0, 0.0, 0.0)); RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(sky->quarter_res_framebuffer, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_DISCARD, clear_colors); - storage->get_effects()->render_sky(draw_list, p_time, sky->quarter_res_framebuffer, sky_scene_state.uniform_set, sky_scene_state.fog_uniform_set, pipeline, material->uniform_set, texture_uniform_set, view_count, projections, sky_transform, multiplier, p_transform.origin); + _render_sky(draw_list, p_time, sky->quarter_res_framebuffer, pipeline, material->uniform_set, texture_uniform_set, view_count, projections, sky_transform, multiplier, p_transform.origin, 1.0); RD::get_singleton()->draw_list_end(); } @@ -1318,7 +1453,7 @@ void RendererSceneSkyRD::draw(RendererSceneEnvironmentRD *p_env, bool p_can_cont clear_colors.push_back(Color(0.0, 0.0, 0.0)); RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(sky->half_res_framebuffer, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_DISCARD, clear_colors); - storage->get_effects()->render_sky(draw_list, p_time, sky->half_res_framebuffer, sky_scene_state.uniform_set, sky_scene_state.fog_uniform_set, pipeline, material->uniform_set, texture_uniform_set, view_count, projections, sky_transform, multiplier, p_transform.origin); + _render_sky(draw_list, p_time, sky->half_res_framebuffer, pipeline, material->uniform_set, texture_uniform_set, view_count, projections, sky_transform, multiplier, p_transform.origin, 1.0); RD::get_singleton()->draw_list_end(); } @@ -1332,10 +1467,183 @@ void RendererSceneSkyRD::draw(RendererSceneEnvironmentRD *p_env, bool p_can_cont } RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(p_fb, RD::INITIAL_ACTION_CONTINUE, p_can_continue_color ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CONTINUE, p_can_continue_depth ? RD::FINAL_ACTION_CONTINUE : RD::FINAL_ACTION_READ); - storage->get_effects()->render_sky(draw_list, p_time, p_fb, sky_scene_state.uniform_set, sky_scene_state.fog_uniform_set, pipeline, material->uniform_set, texture_uniform_set, view_count, projections, sky_transform, multiplier, p_transform.origin); + _render_sky(draw_list, p_time, p_fb, pipeline, material->uniform_set, texture_uniform_set, view_count, projections, sky_transform, multiplier, p_transform.origin, 1.0); RD::get_singleton()->draw_list_end(); } +void RendererSceneSkyRD::update_res_buffers(RendererSceneEnvironmentRD *p_env, uint32_t p_view_count, const CameraMatrix *p_projections, const Transform3D &p_transform, double p_time, float p_luminance_multiplier) { + ERR_FAIL_COND(!p_env); + + ERR_FAIL_COND(p_view_count == 0); + ERR_FAIL_COND(p_view_count > RendererSceneRender::MAX_RENDER_VIEWS); + + Sky *sky = get_sky(p_env->sky); + ERR_FAIL_COND(!sky); + + SkyMaterialData *material = nullptr; + RID sky_material; + + RS::EnvironmentBG background = p_env->background; + + if (!(background == RS::ENV_BG_CLEAR_COLOR || background == RS::ENV_BG_COLOR) || sky) { + ERR_FAIL_COND(!sky); + sky_material = sky_get_material(p_env->sky); + + if (sky_material.is_valid()) { + material = (SkyMaterialData *)storage->material_get_data(sky_material, RendererStorageRD::SHADER_TYPE_SKY); + if (!material || !material->shader_data->valid) { + material = nullptr; + } + } + + if (!material) { + sky_material = sky_shader.default_material; + material = (SkyMaterialData *)storage->material_get_data(sky_material, RendererStorageRD::SHADER_TYPE_SKY); + } + } + + if (background == RS::ENV_BG_CLEAR_COLOR || background == RS::ENV_BG_COLOR) { + sky_material = sky_scene_state.fog_material; + material = (SkyMaterialData *)storage->material_get_data(sky_material, RendererStorageRD::SHADER_TYPE_SKY); + } + + ERR_FAIL_COND(!material); + + SkyShaderData *shader_data = material->shader_data; + + ERR_FAIL_COND(!shader_data); + + Basis sky_transform = p_env->sky_orientation; + sky_transform.invert(); + + float multiplier = p_env->bg_energy; + float custom_fov = p_env->sky_custom_fov; + + // Camera + CameraMatrix camera; + uint32_t view_count = p_view_count; + const CameraMatrix *projections = p_projections; + + if (custom_fov) { + // With custom fov we don't support stereo... + float near_plane = p_projections[0].get_z_near(); + float far_plane = p_projections[0].get_z_far(); + float aspect = p_projections[0].get_aspect(); + + camera.set_perspective(custom_fov, aspect, near_plane, far_plane); + + view_count = 1; + projections = &camera; + } + + sky_transform = p_transform.basis * sky_transform; + + if (shader_data->uses_quarter_res) { + PipelineCacheRD *pipeline = &shader_data->pipelines[view_count > 1 ? SKY_VERSION_QUARTER_RES_MULTIVIEW : SKY_VERSION_QUARTER_RES]; + + RID texture_uniform_set = sky->get_textures(storage, SKY_TEXTURE_SET_QUARTER_RES, sky_shader.default_shader_rd); + + Vector<Color> clear_colors; + clear_colors.push_back(Color(0.0, 0.0, 0.0)); + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(sky->quarter_res_framebuffer, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_DISCARD, clear_colors); + _render_sky(draw_list, p_time, sky->quarter_res_framebuffer, pipeline, material->uniform_set, texture_uniform_set, view_count, projections, sky_transform, multiplier, p_transform.origin, p_luminance_multiplier); + RD::get_singleton()->draw_list_end(); + } + + if (shader_data->uses_half_res) { + PipelineCacheRD *pipeline = &shader_data->pipelines[view_count > 1 ? SKY_VERSION_HALF_RES_MULTIVIEW : SKY_VERSION_HALF_RES]; + + RID texture_uniform_set = sky->get_textures(storage, SKY_TEXTURE_SET_HALF_RES, sky_shader.default_shader_rd); + + Vector<Color> clear_colors; + clear_colors.push_back(Color(0.0, 0.0, 0.0)); + + RD::DrawListID draw_list = RD::get_singleton()->draw_list_begin(sky->half_res_framebuffer, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_READ, RD::INITIAL_ACTION_CLEAR, RD::FINAL_ACTION_DISCARD, clear_colors); + _render_sky(draw_list, p_time, sky->half_res_framebuffer, pipeline, material->uniform_set, texture_uniform_set, view_count, projections, sky_transform, multiplier, p_transform.origin, p_luminance_multiplier); + RD::get_singleton()->draw_list_end(); + } +} + +void RendererSceneSkyRD::draw(RD::DrawListID p_draw_list, RendererSceneEnvironmentRD *p_env, RID p_fb, uint32_t p_view_count, const CameraMatrix *p_projections, const Transform3D &p_transform, double p_time, float p_luminance_multiplier) { + ERR_FAIL_COND(!p_env); + + ERR_FAIL_COND(p_view_count == 0); + ERR_FAIL_COND(p_view_count > RendererSceneRender::MAX_RENDER_VIEWS); + + Sky *sky = get_sky(p_env->sky); + ERR_FAIL_COND(!sky); + + SkyMaterialData *material = nullptr; + RID sky_material; + + RS::EnvironmentBG background = p_env->background; + + if (!(background == RS::ENV_BG_CLEAR_COLOR || background == RS::ENV_BG_COLOR) || sky) { + ERR_FAIL_COND(!sky); + sky_material = sky_get_material(p_env->sky); + + if (sky_material.is_valid()) { + material = (SkyMaterialData *)storage->material_get_data(sky_material, RendererStorageRD::SHADER_TYPE_SKY); + if (!material || !material->shader_data->valid) { + material = nullptr; + } + } + + if (!material) { + sky_material = sky_shader.default_material; + material = (SkyMaterialData *)storage->material_get_data(sky_material, RendererStorageRD::SHADER_TYPE_SKY); + } + } + + if (background == RS::ENV_BG_CLEAR_COLOR || background == RS::ENV_BG_COLOR) { + sky_material = sky_scene_state.fog_material; + material = (SkyMaterialData *)storage->material_get_data(sky_material, RendererStorageRD::SHADER_TYPE_SKY); + } + + ERR_FAIL_COND(!material); + + SkyShaderData *shader_data = material->shader_data; + + ERR_FAIL_COND(!shader_data); + + Basis sky_transform = p_env->sky_orientation; + sky_transform.invert(); + + float multiplier = p_env->bg_energy; + float custom_fov = p_env->sky_custom_fov; + + // Camera + CameraMatrix camera; + uint32_t view_count = p_view_count; + const CameraMatrix *projections = p_projections; + + if (custom_fov) { + // With custom fov we don't support stereo... + float near_plane = p_projections[0].get_z_near(); + float far_plane = p_projections[0].get_z_far(); + float aspect = p_projections[0].get_aspect(); + + camera.set_perspective(custom_fov, aspect, near_plane, far_plane); + + view_count = 1; + projections = &camera; + } + + sky_transform = p_transform.basis * sky_transform; + + PipelineCacheRD *pipeline = &shader_data->pipelines[view_count > 1 ? SKY_VERSION_BACKGROUND_MULTIVIEW : SKY_VERSION_BACKGROUND]; + + RID texture_uniform_set; + if (sky) { + texture_uniform_set = sky->get_textures(storage, SKY_TEXTURE_SET_BACKGROUND, sky_shader.default_shader_rd); + } else { + texture_uniform_set = sky_scene_state.fog_only_texture_uniform_set; + } + + _render_sky(p_draw_list, p_time, p_fb, pipeline, material->uniform_set, texture_uniform_set, view_count, projections, sky_transform, multiplier, p_transform.origin, p_luminance_multiplier); +} + void RendererSceneSkyRD::invalidate_sky(Sky *p_sky) { if (!p_sky->dirty) { p_sky->dirty = true; @@ -1367,7 +1675,7 @@ void RendererSceneSkyRD::update_dirty_skys() { //array (higher quality, 6 times more memory) RD::TextureFormat tf; tf.array_layers = layers * 6; - tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + tf.format = texture_format; tf.texture_type = RD::TEXTURE_TYPE_CUBE_ARRAY; tf.mipmaps = mipmaps; tf.width = w; @@ -1376,13 +1684,13 @@ void RendererSceneSkyRD::update_dirty_skys() { sky->radiance = RD::get_singleton()->texture_create(tf, RD::TextureView()); - sky->reflection.update_reflection_data(sky->radiance_size, mipmaps, true, sky->radiance, 0, sky->mode == RS::SKY_MODE_REALTIME, roughness_layers); + sky->reflection.update_reflection_data(storage, sky->radiance_size, mipmaps, true, sky->radiance, 0, sky->mode == RS::SKY_MODE_REALTIME, roughness_layers, texture_format); } else { //regular cubemap, lower quality (aliasing, less memory) RD::TextureFormat tf; tf.array_layers = 6; - tf.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + tf.format = texture_format; tf.texture_type = RD::TEXTURE_TYPE_CUBE; tf.mipmaps = MIN(mipmaps, layers); tf.width = w; @@ -1391,7 +1699,7 @@ void RendererSceneSkyRD::update_dirty_skys() { sky->radiance = RD::get_singleton()->texture_create(tf, RD::TextureView()); - sky->reflection.update_reflection_data(sky->radiance_size, MIN(mipmaps, layers), false, sky->radiance, 0, sky->mode == RS::SKY_MODE_REALTIME, roughness_layers); + sky->reflection.update_reflection_data(storage, sky->radiance_size, MIN(mipmaps, layers), false, sky->radiance, 0, sky->mode == RS::SKY_MODE_REALTIME, roughness_layers, texture_format); } texture_set_dirty = true; } @@ -1399,7 +1707,7 @@ void RendererSceneSkyRD::update_dirty_skys() { // Create subpass buffers if they haven't been created already if (sky->half_res_pass.is_null() && !RD::get_singleton()->texture_is_valid(sky->half_res_pass) && sky->screen_size.x >= 4 && sky->screen_size.y >= 4) { RD::TextureFormat tformat; - tformat.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + tformat.format = texture_format; tformat.width = sky->screen_size.x / 2; tformat.height = sky->screen_size.y / 2; tformat.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; @@ -1414,7 +1722,7 @@ void RendererSceneSkyRD::update_dirty_skys() { if (sky->quarter_res_pass.is_null() && !RD::get_singleton()->texture_is_valid(sky->quarter_res_pass) && sky->screen_size.x >= 4 && sky->screen_size.y >= 4) { RD::TextureFormat tformat; - tformat.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + tformat.format = texture_format; tformat.width = sky->screen_size.x / 4; tformat.height = sky->screen_size.y / 4; tformat.usage_bits = RD::TEXTURE_USAGE_SAMPLING_BIT | RD::TEXTURE_USAGE_COLOR_ATTACHMENT_BIT; diff --git a/servers/rendering/renderer_rd/renderer_scene_sky_rd.h b/servers/rendering/renderer_rd/renderer_scene_sky_rd.h index 200902bff2..7f563c9bc4 100644 --- a/servers/rendering/renderer_rd/renderer_scene_sky_rd.h +++ b/servers/rendering/renderer_rd/renderer_scene_sky_rd.h @@ -43,9 +43,6 @@ class RendererSceneRenderRD; class RendererSceneSkyRD { -private: - RendererStorageRD *storage; - public: enum SkySet { SKY_SET_UNIFORMS, @@ -55,6 +52,23 @@ public: SKY_SET_MAX }; + // Skys need less info from Directional Lights than the normal shaders + struct SkyDirectionalLightData { + float direction[3]; + float energy; + float color[3]; + float size; + uint32_t enabled; + uint32_t pad[3]; + }; + +private: + RendererStorageRD *storage; + RD::DataFormat texture_format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; + + RID index_buffer; + RID index_array; + enum SkyTextureSetVersion { SKY_TEXTURE_SET_BACKGROUND, SKY_TEXTURE_SET_HALF_RES, @@ -80,16 +94,54 @@ public: SKY_VERSION_MAX }; - // Skys need less info from Directional Lights than the normal shaders - struct SkyDirectionalLightData { - float direction[3]; - float energy; - float color[3]; - float size; - uint32_t enabled; - uint32_t pad[3]; + struct SkyPushConstant { + float orientation[12]; // 48 - 48 + float projections[RendererSceneRender::MAX_RENDER_VIEWS][4]; // 2 x 16 - 80 + float position[3]; // 12 - 92 + float multiplier; // 4 - 96 + float time; // 4 - 100 + float luminance_multiplier; // 4 - 104 + float pad[2]; // 8 - 112 // Using pad to align on 16 bytes + // 128 is the max size of a push constant. We can replace "pad" but we can't add any more. + }; + + struct SkyShaderData : public RendererStorageRD::ShaderData { + bool valid; + RID version; + + PipelineCacheRD pipelines[SKY_VERSION_MAX]; + Map<StringName, ShaderLanguage::ShaderNode::Uniform> uniforms; + Vector<ShaderCompilerRD::GeneratedCode::Texture> texture_uniforms; + + Vector<uint32_t> ubo_offsets; + uint32_t ubo_size; + + String path; + String code; + Map<StringName, RID> default_texture_params; + + bool uses_time; + bool uses_position; + bool uses_half_res; + bool uses_quarter_res; + bool uses_light; + + virtual void set_code(const String &p_Code); + virtual void set_default_texture_param(const StringName &p_name, RID p_texture); + virtual void get_param_list(List<PropertyInfo> *p_param_list) const; + virtual void get_instance_param_list(List<RendererStorage::InstanceShaderParam> *p_param_list) const; + virtual bool is_param_texture(const StringName &p_param) const; + virtual bool is_animated() const; + virtual bool casts_shadows() const; + virtual Variant get_default_parameter(const StringName &p_parameter) const; + virtual RS::ShaderNativeSourceCode get_native_source_code() const; + SkyShaderData(); + virtual ~SkyShaderData(); }; + void _render_sky(RD::DrawListID p_list, float p_time, RID p_fb, PipelineCacheRD *p_pipeline, RID p_uniform_set, RID p_texture_set, uint32_t p_view_count, const CameraMatrix *p_projections, const Basis &p_orientation, float p_multiplier, const Vector3 &p_position, float p_luminance_multiplier); + +public: struct SkySceneState { struct UBO { uint32_t volumetric_fog_enabled; @@ -140,6 +192,10 @@ public: struct Mipmap { RID view; Size2i size; + + // for mobile only + RID views[6]; + RID framebuffers[6]; }; Vector<Mipmap> mipmaps; }; @@ -154,46 +210,12 @@ public: Vector<Layer> layers; void clear_reflection_data(); - void update_reflection_data(int p_size, int p_mipmaps, bool p_use_array, RID p_base_cube, int p_base_layer, bool p_low_quality, int p_roughness_layers); + void update_reflection_data(RendererStorageRD *p_storage, int p_size, int p_mipmaps, bool p_use_array, RID p_base_cube, int p_base_layer, bool p_low_quality, int p_roughness_layers, RD::DataFormat p_texture_format); void create_reflection_fast_filter(RendererStorageRD *p_storage, bool p_use_arrays); void create_reflection_importance_sample(RendererStorageRD *p_storage, bool p_use_arrays, int p_cube_side, int p_base_layer, uint32_t p_sky_ggx_samples_quality); void update_reflection_mipmaps(RendererStorageRD *p_storage, int p_start, int p_end); }; - struct SkyShaderData : public RendererStorageRD::ShaderData { - bool valid; - RID version; - - PipelineCacheRD pipelines[SKY_VERSION_MAX]; - Map<StringName, ShaderLanguage::ShaderNode::Uniform> uniforms; - Vector<ShaderCompilerRD::GeneratedCode::Texture> texture_uniforms; - - Vector<uint32_t> ubo_offsets; - uint32_t ubo_size; - - String path; - String code; - Map<StringName, RID> default_texture_params; - - bool uses_time; - bool uses_position; - bool uses_half_res; - bool uses_quarter_res; - bool uses_light; - - virtual void set_code(const String &p_Code); - virtual void set_default_texture_param(const StringName &p_name, RID p_texture); - virtual void get_param_list(List<PropertyInfo> *p_param_list) const; - virtual void get_instance_param_list(List<RendererStorage::InstanceShaderParam> *p_param_list) const; - virtual bool is_param_texture(const StringName &p_param) const; - virtual bool is_animated() const; - virtual bool casts_shadows() const; - virtual Variant get_default_parameter(const StringName &p_parameter) const; - virtual RS::ShaderNativeSourceCode get_native_source_code() const; - SkyShaderData(); - virtual ~SkyShaderData(); - }; - /* Sky shader */ struct SkyShader { @@ -208,15 +230,12 @@ public: struct SkyMaterialData : public RendererStorageRD::MaterialData { uint64_t last_frame; SkyShaderData *shader_data; - RID uniform_buffer; RID uniform_set; - Vector<RID> texture_cache; - Vector<uint8_t> ubo_data; bool uniform_set_updated; virtual void set_render_priority(int p_priority) {} virtual void set_next_pass(RID p_pass) {} - virtual void update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty); + virtual bool update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty); virtual ~SkyMaterialData(); }; @@ -270,12 +289,15 @@ public: static RendererStorageRD::MaterialData *_create_sky_material_funcs(RendererStorageRD::ShaderData *p_shader); RendererSceneSkyRD(); - void init(RendererStorageRD *p_storage); + void set_texture_format(RD::DataFormat p_texture_format); + ~RendererSceneSkyRD(); void setup(RendererSceneEnvironmentRD *p_env, RID p_render_buffers, const CameraMatrix &p_projection, const Transform3D &p_transform, const Size2i p_screen_size, RendererSceneRenderRD *p_scene_render); - void update(RendererSceneEnvironmentRD *p_env, const CameraMatrix &p_projection, const Transform3D &p_transform, double p_time); - void draw(RendererSceneEnvironmentRD *p_env, bool p_can_continue_color, bool p_can_continue_depth, RID p_fb, uint32_t p_view_count, const CameraMatrix *p_projections, const Transform3D &p_transform, double p_time); + void update(RendererSceneEnvironmentRD *p_env, const CameraMatrix &p_projection, const Transform3D &p_transform, double p_time, float p_luminance_multiplier = 1.0); + void draw(RendererSceneEnvironmentRD *p_env, bool p_can_continue_color, bool p_can_continue_depth, RID p_fb, uint32_t p_view_count, const CameraMatrix *p_projections, const Transform3D &p_transform, double p_time); // only called by clustered renderer + void update_res_buffers(RendererSceneEnvironmentRD *p_env, uint32_t p_view_count, const CameraMatrix *p_projections, const Transform3D &p_transform, double p_time, float p_luminance_multiplier = 1.0); + void draw(RD::DrawListID p_draw_list, RendererSceneEnvironmentRD *p_env, RID p_fb, uint32_t p_view_count, const CameraMatrix *p_projections, const Transform3D &p_transform, double p_time, float p_luminance_multiplier = 1.0); void invalidate_sky(Sky *p_sky); void update_dirty_skys(); diff --git a/servers/rendering/renderer_rd/renderer_storage_rd.cpp b/servers/rendering/renderer_rd/renderer_storage_rd.cpp index eeca283c13..ec0d25376f 100644 --- a/servers/rendering/renderer_rd/renderer_storage_rd.cpp +++ b/servers/rendering/renderer_rd/renderer_storage_rd.cpp @@ -35,6 +35,7 @@ #include "core/io/resource_loader.h" #include "core/math/math_defs.h" #include "renderer_compositor_rd.h" +#include "servers/rendering/rendering_server_globals.h" #include "servers/rendering/shader_language.h" bool RendererStorageRD::can_create_resources_async() const { @@ -883,10 +884,6 @@ void RendererStorageRD::_texture_2d_update(RID p_texture, const Ref<Image> &p_im RD::get_singleton()->texture_update(tex->rd_texture, p_layer, validated->get_data()); } -void RendererStorageRD::texture_2d_update_immediate(RID p_texture, const Ref<Image> &p_image, int p_layer) { - _texture_2d_update(p_texture, p_image, p_layer, true); -} - void RendererStorageRD::texture_2d_update(RID p_texture, const Ref<Image> &p_image, int p_layer) { _texture_2d_update(p_texture, p_image, p_layer, false); } @@ -972,7 +969,7 @@ void RendererStorageRD::texture_2d_placeholder_initialize(RID p_texture) { //this could be better optimized to reuse an existing image , done this way //for now to get it working Ref<Image> image; - image.instance(); + image.instantiate(); image->create(4, 4, false, Image::FORMAT_RGBA8); for (int i = 0; i < 4; i++) { @@ -988,7 +985,7 @@ void RendererStorageRD::texture_2d_layered_placeholder_initialize(RID p_texture, //this could be better optimized to reuse an existing image , done this way //for now to get it working Ref<Image> image; - image.instance(); + image.instantiate(); image->create(4, 4, false, Image::FORMAT_RGBA8); for (int i = 0; i < 4; i++) { @@ -1014,7 +1011,7 @@ void RendererStorageRD::texture_3d_placeholder_initialize(RID p_texture) { //this could be better optimized to reuse an existing image , done this way //for now to get it working Ref<Image> image; - image.instance(); + image.instantiate(); image->create(4, 4, false, Image::FORMAT_RGBA8); for (int i = 0; i < 4; i++) { @@ -1044,7 +1041,7 @@ Ref<Image> RendererStorageRD::texture_2d_get(RID p_texture) const { Vector<uint8_t> data = RD::get_singleton()->texture_get_data(tex->rd_texture, 0); ERR_FAIL_COND_V(data.size() == 0, Ref<Image>()); Ref<Image> image; - image.instance(); + image.instantiate(); image->create(tex->width, tex->height, tex->mipmaps > 1, tex->validated_format, data); ERR_FAIL_COND_V(image->is_empty(), Ref<Image>()); if (tex->format != tex->validated_format) { @@ -1067,7 +1064,7 @@ Ref<Image> RendererStorageRD::texture_2d_layer_get(RID p_texture, int p_layer) c Vector<uint8_t> data = RD::get_singleton()->texture_get_data(tex->rd_texture, p_layer); ERR_FAIL_COND_V(data.size() == 0, Ref<Image>()); Ref<Image> image; - image.instance(); + image.instantiate(); image->create(tex->width, tex->height, tex->mipmaps > 1, tex->validated_format, data); ERR_FAIL_COND_V(image->is_empty(), Ref<Image>()); if (tex->format != tex->validated_format) { @@ -1095,7 +1092,7 @@ Vector<Ref<Image>> RendererStorageRD::texture_3d_get(RID p_texture) const { Vector<uint8_t> sub_region = all_data.subarray(bs.offset, bs.offset + bs.buffer_size - 1); Ref<Image> img; - img.instance(); + img.instantiate(); img->create(bs.size.width, bs.size.height, false, tex->validated_format, sub_region); ERR_FAIL_COND_V(img->is_empty(), Vector<Ref<Image>>()); if (tex->format != tex->validated_format) { @@ -1234,7 +1231,7 @@ RID RendererStorageRD::canvas_texture_allocate() { return canvas_texture_owner.allocate_rid(); } void RendererStorageRD::canvas_texture_initialize(RID p_rid) { - canvas_texture_owner.initialize_rid(p_rid, memnew(CanvasTexture)); + canvas_texture_owner.initialize_rid(p_rid); } void RendererStorageRD::canvas_texture_set_channel(RID p_canvas_texture, RS::CanvasTextureChannel p_channel, RID p_texture) { @@ -1440,8 +1437,10 @@ void RendererStorageRD::shader_set_code(RID p_shader, const String &p_code) { material->shader_type = new_type; } - for (Map<StringName, RID>::Element *E = shader->default_texture_parameter.front(); E; E = E->next()) { - shader->data->set_default_texture_param(E->key(), E->get()); + if (shader->data) { + for (Map<StringName, RID>::Element *E = shader->default_texture_parameter.front(); E; E = E->next()) { + shader->data->set_default_texture_param(E->key(), E->get()); + } } } @@ -1527,27 +1526,18 @@ RID RendererStorageRD::material_allocate() { return material_owner.allocate_rid(); } void RendererStorageRD::material_initialize(RID p_rid) { - Material material; - material.data = nullptr; - material.shader = nullptr; - material.shader_type = SHADER_TYPE_MAX; - material.update_next = nullptr; - material.update_requested = false; - material.uniform_dirty = false; - material.texture_dirty = false; - material.priority = 0; - material.self = p_rid; - material_owner.initialize_rid(p_rid, material); + material_owner.initialize_rid(p_rid); + Material *material = material_owner.getornull(p_rid); + material->self = p_rid; } void RendererStorageRD::_material_queue_update(Material *material, bool p_uniform, bool p_texture) { - if (material->update_requested) { + if (material->update_element.in_list()) { return; } - material->update_next = material_update_list; - material_update_list = material; - material->update_requested = true; + material_update_list.add(&material->update_element); + material->uniform_dirty = material->uniform_dirty || p_uniform; material->texture_dirty = material->texture_dirty || p_texture; } @@ -1602,6 +1592,7 @@ void RendererStorageRD::material_set_param(RID p_material, const StringName &p_p if (p_value.get_type() == Variant::NIL) { material->params.erase(p_param); } else { + ERR_FAIL_COND(p_value.get_type() == Variant::OBJECT); //object not allowed material->params[p_param] = p_value; } @@ -2233,6 +2224,10 @@ RendererStorageRD::MaterialData::~MaterialData() { //unregister material from those using global textures rs->global_variables.materials_using_texture.erase(global_texture_E); } + + if (uniform_buffer.is_valid()) { + RD::get_singleton()->free(uniform_buffer); + } } void RendererStorageRD::MaterialData::update_textures(const Map<StringName, Variant> &p_parameters, const Map<StringName, RID> &p_default_textures, const Vector<ShaderCompilerRD::GeneratedCode::Texture> &p_texture_uniforms, RID *p_textures, bool p_use_linear_color) { @@ -2382,6 +2377,105 @@ void RendererStorageRD::MaterialData::update_textures(const Map<StringName, Vari } } +void RendererStorageRD::MaterialData::free_parameters_uniform_set(RID p_uniform_set) { + if (p_uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(p_uniform_set)) { + RD::get_singleton()->uniform_set_set_invalidation_callback(p_uniform_set, nullptr, nullptr); + RD::get_singleton()->free(p_uniform_set); + } +} + +bool RendererStorageRD::MaterialData::update_parameters_uniform_set(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty, const Map<StringName, ShaderLanguage::ShaderNode::Uniform> &p_uniforms, const uint32_t *p_uniform_offsets, const Vector<ShaderCompilerRD::GeneratedCode::Texture> &p_texture_uniforms, const Map<StringName, RID> &p_default_texture_params, uint32_t p_ubo_size, RID &uniform_set, RID p_shader, uint32_t p_shader_uniform_set, uint32_t p_barrier) { + if ((uint32_t)ubo_data.size() != p_ubo_size) { + p_uniform_dirty = true; + if (uniform_buffer.is_valid()) { + RD::get_singleton()->free(uniform_buffer); + uniform_buffer = RID(); + } + + ubo_data.resize(p_ubo_size); + if (ubo_data.size()) { + uniform_buffer = RD::get_singleton()->uniform_buffer_create(ubo_data.size()); + memset(ubo_data.ptrw(), 0, ubo_data.size()); //clear + } + + //clear previous uniform set + if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { + RD::get_singleton()->uniform_set_set_invalidation_callback(uniform_set, nullptr, nullptr); + RD::get_singleton()->free(uniform_set); + uniform_set = RID(); + } + } + + //check whether buffer changed + if (p_uniform_dirty && ubo_data.size()) { + update_uniform_buffer(p_uniforms, p_uniform_offsets, p_parameters, ubo_data.ptrw(), ubo_data.size(), false); + RD::get_singleton()->buffer_update(uniform_buffer, 0, ubo_data.size(), ubo_data.ptrw(), p_barrier); + } + + uint32_t tex_uniform_count = p_texture_uniforms.size(); + + if ((uint32_t)texture_cache.size() != tex_uniform_count || p_textures_dirty) { + texture_cache.resize(tex_uniform_count); + p_textures_dirty = true; + + //clear previous uniform set + if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { + RD::get_singleton()->uniform_set_set_invalidation_callback(uniform_set, nullptr, nullptr); + RD::get_singleton()->free(uniform_set); + uniform_set = RID(); + } + } + + if (p_textures_dirty && tex_uniform_count) { + update_textures(p_parameters, p_default_texture_params, p_texture_uniforms, texture_cache.ptrw(), true); + } + + if (p_ubo_size == 0 && p_texture_uniforms.size() == 0) { + // This material does not require an uniform set, so don't create it. + return false; + } + + if (!p_textures_dirty && uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { + //no reason to update uniform set, only UBO (or nothing) was needed to update + return false; + } + + Vector<RD::Uniform> uniforms; + + { + if (p_ubo_size) { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; + u.binding = 0; + u.ids.push_back(uniform_buffer); + uniforms.push_back(u); + } + + const RID *textures = texture_cache.ptrw(); + for (uint32_t i = 0; i < tex_uniform_count; i++) { + RD::Uniform u; + u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; + u.binding = 1 + i; + u.ids.push_back(textures[i]); + uniforms.push_back(u); + } + } + + uniform_set = RD::get_singleton()->uniform_set_create(uniforms, p_shader, p_shader_uniform_set); + + RD::get_singleton()->uniform_set_set_invalidation_callback(uniform_set, _material_uniform_set_erased, &self); + + return true; +} + +void RendererStorageRD::_material_uniform_set_erased(const RID &p_set, void *p_material) { + RID rid = *(RID *)p_material; + Material *material = base_singleton->material_owner.getornull(rid); + if (material) { + material->dependency.changed_notify(DEPENDENCY_CHANGED_MATERIAL); + } +} + void RendererStorageRD::material_force_update_textures(RID p_material, ShaderType p_shader_type) { Material *material = material_owner.getornull(p_material); if (material->shader_type != p_shader_type) { @@ -2393,20 +2487,23 @@ void RendererStorageRD::material_force_update_textures(RID p_material, ShaderTyp } void RendererStorageRD::_update_queued_materials() { - Material *material = material_update_list; - while (material) { - Material *next = material->update_next; + while (material_update_list.first()) { + Material *material = material_update_list.first()->self(); + bool uniforms_changed = false; if (material->data) { - material->data->update_parameters(material->params, material->uniform_dirty, material->texture_dirty); + uniforms_changed = material->data->update_parameters(material->params, material->uniform_dirty, material->texture_dirty); } - material->update_requested = false; material->texture_dirty = false; material->uniform_dirty = false; - material->update_next = nullptr; - material = next; + + material_update_list.remove(&material->update_element); + + if (uniforms_changed) { + //some implementations such as 3D renderer cache the matreial uniform set, so update is required + material->dependency.changed_notify(DEPENDENCY_CHANGED_MATERIAL); + } } - material_update_list = nullptr; } /* MESH API */ @@ -2434,6 +2531,8 @@ void RendererStorageRD::mesh_add_surface(RID p_mesh, const RS::SurfaceData &p_su Mesh *mesh = mesh_owner.getornull(p_mesh); ERR_FAIL_COND(!mesh); + ERR_FAIL_COND(mesh->surface_count == RS::MAX_MESH_SURFACES); + #ifdef DEBUG_ENABLED //do a validation, to catch errors first { @@ -2461,7 +2560,7 @@ void RendererStorageRD::mesh_add_surface(RID p_mesh, const RS::SurfaceData &p_su } break; case RS::ARRAY_COLOR: { - attrib_stride += sizeof(int16_t) * 4; + attrib_stride += sizeof(uint32_t); } break; case RS::ARRAY_TEX_UV: { attrib_stride += sizeof(float) * 2; @@ -2549,6 +2648,7 @@ void RendererStorageRD::mesh_add_surface(RID p_mesh, const RS::SurfaceData &p_su s->lods[i].index_buffer = RD::get_singleton()->index_buffer_create(indices, is_index_16 ? RD::INDEX_BUFFER_FORMAT_UINT16 : RD::INDEX_BUFFER_FORMAT_UINT32, p_surface.lods[i].index_data); s->lods[i].index_array = RD::get_singleton()->index_array_create(s->lods[i].index_buffer, 0, indices); s->lods[i].edge_length = p_surface.lods[i].edge_length; + s->lods[i].index_count = indices; } } } @@ -2616,9 +2716,7 @@ void RendererStorageRD::mesh_add_surface(RID p_mesh, const RS::SurfaceData &p_su mesh->surfaces[mesh->surface_count] = s; mesh->surface_count++; - for (List<MeshInstance *>::Element *E = mesh->instances.front(); E; E = E->next()) { - //update instances - MeshInstance *mi = E->get(); + for (MeshInstance *mi : mesh->instances) { _mesh_instance_add_surface(mi, mesh, mesh->surface_count - 1); } @@ -2653,7 +2751,7 @@ RS::BlendShapeMode RendererStorageRD::mesh_get_blend_shape_mode(RID p_mesh) cons return mesh->blend_shape_mode; } -void RendererStorageRD::mesh_surface_update_region(RID p_mesh, int p_surface, int p_offset, const Vector<uint8_t> &p_data) { +void RendererStorageRD::mesh_surface_update_vertex_region(RID p_mesh, int p_surface, int p_offset, const Vector<uint8_t> &p_data) { Mesh *mesh = mesh_owner.getornull(p_mesh); ERR_FAIL_COND(!mesh); ERR_FAIL_UNSIGNED_INDEX((uint32_t)p_surface, mesh->surface_count); @@ -2664,6 +2762,30 @@ void RendererStorageRD::mesh_surface_update_region(RID p_mesh, int p_surface, in RD::get_singleton()->buffer_update(mesh->surfaces[p_surface]->vertex_buffer, p_offset, data_size, r); } +void RendererStorageRD::mesh_surface_update_attribute_region(RID p_mesh, int p_surface, int p_offset, const Vector<uint8_t> &p_data) { + Mesh *mesh = mesh_owner.getornull(p_mesh); + ERR_FAIL_COND(!mesh); + ERR_FAIL_UNSIGNED_INDEX((uint32_t)p_surface, mesh->surface_count); + ERR_FAIL_COND(p_data.size() == 0); + ERR_FAIL_COND(mesh->surfaces[p_surface]->attribute_buffer.is_null()); + uint64_t data_size = p_data.size(); + const uint8_t *r = p_data.ptr(); + + RD::get_singleton()->buffer_update(mesh->surfaces[p_surface]->attribute_buffer, p_offset, data_size, r); +} + +void RendererStorageRD::mesh_surface_update_skin_region(RID p_mesh, int p_surface, int p_offset, const Vector<uint8_t> &p_data) { + Mesh *mesh = mesh_owner.getornull(p_mesh); + ERR_FAIL_COND(!mesh); + ERR_FAIL_UNSIGNED_INDEX((uint32_t)p_surface, mesh->surface_count); + ERR_FAIL_COND(p_data.size() == 0); + ERR_FAIL_COND(mesh->surfaces[p_surface]->skin_buffer.is_null()); + uint64_t data_size = p_data.size(); + const uint8_t *r = p_data.ptr(); + + RD::get_singleton()->buffer_update(mesh->surfaces[p_surface]->skin_buffer, p_offset, data_size, r); +} + void RendererStorageRD::mesh_surface_set_material(RID p_mesh, int p_surface, RID p_material) { Mesh *mesh = mesh_owner.getornull(p_mesh); ERR_FAIL_COND(!mesh); @@ -2905,8 +3027,7 @@ void RendererStorageRD::mesh_clear(RID p_mesh) { mesh->surface_count = 0; mesh->material_cache.clear(); //clear instance data - for (List<MeshInstance *>::Element *E = mesh->instances.front(); E; E = E->next()) { - MeshInstance *mi = E->get(); + for (MeshInstance *mi : mesh->instances) { _mesh_instance_clear(mi); } mesh->has_bone_weights = false; @@ -2932,7 +3053,8 @@ RID RendererStorageRD::mesh_instance_create(RID p_base) { Mesh *mesh = mesh_owner.getornull(p_base); ERR_FAIL_COND_V(!mesh, RID()); - MeshInstance *mi = memnew(MeshInstance); + RID rid = mesh_instance_owner.make_rid(); + MeshInstance *mi = mesh_instance_owner.getornull(rid); mi->mesh = mesh; @@ -2944,7 +3066,7 @@ RID RendererStorageRD::mesh_instance_create(RID p_base) { mi->dirty = true; - return mesh_instance_owner.make_rid(mi); + return rid; } void RendererStorageRD::mesh_instance_set_skeleton(RID p_mesh_instance, RID p_skeleton) { MeshInstance *mi = mesh_instance_owner.getornull(p_mesh_instance); @@ -3231,8 +3353,8 @@ void RendererStorageRD::_mesh_surface_generate_version_for_input_mask(Mesh::Surf case RS::ARRAY_COLOR: { vd.offset = attribute_stride; - vd.format = RD::DATA_FORMAT_R16G16B16A16_SFLOAT; - attribute_stride += sizeof(int16_t) * 4; + vd.format = RD::DATA_FORMAT_R8G8B8A8_UNORM; + attribute_stride += sizeof(int8_t) * 4; buffer = s->attribute_buffer; } break; case RS::ARRAY_TEX_UV: { @@ -3331,6 +3453,7 @@ void RendererStorageRD::multimesh_allocate_data(RID p_multimesh, int p_instances if (multimesh->buffer.is_valid()) { RD::get_singleton()->free(multimesh->buffer); multimesh->buffer = RID(); + multimesh->uniform_set_2d = RID(); //cleared by dependency multimesh->uniform_set_3d = RID(); //cleared by dependency } @@ -3823,7 +3946,7 @@ void RendererStorageRD::_update_dirty_multimeshes() { if (multimesh->data_cache_used_dirty_regions) { uint32_t data_cache_dirty_region_count = (multimesh->instances - 1) / MULTIMESH_DIRTY_REGION_SIZE + 1; - uint32_t visible_region_count = (visible_instances - 1) / MULTIMESH_DIRTY_REGION_SIZE + 1; + uint32_t visible_region_count = visible_instances == 0 ? 0 : (visible_instances - 1) / MULTIMESH_DIRTY_REGION_SIZE + 1; uint32_t region_size = multimesh->stride_cache * MULTIMESH_DIRTY_REGION_SIZE * sizeof(float); @@ -3894,6 +4017,7 @@ void RendererStorageRD::particles_set_emitting(RID p_particles, bool p_emitting) } bool RendererStorageRD::particles_get_emitting(RID p_particles) { + ERR_FAIL_COND_V_MSG(RSG::threaded, false, "This function should never be used with threaded rendering, as it stalls the renderer."); Particles *particles = particles_owner.getornull(p_particles); ERR_FAIL_COND_V(!particles, false); @@ -3968,7 +4092,7 @@ void RendererStorageRD::particles_set_amount(RID p_particles, int p_amount) { particles->dependency.changed_notify(DEPENDENCY_CHANGED_PARTICLES); } -void RendererStorageRD::particles_set_lifetime(RID p_particles, float p_lifetime) { +void RendererStorageRD::particles_set_lifetime(RID p_particles, double p_lifetime) { Particles *particles = particles_owner.getornull(p_particles); ERR_FAIL_COND(!particles); particles->lifetime = p_lifetime; @@ -3980,17 +4104,17 @@ void RendererStorageRD::particles_set_one_shot(RID p_particles, bool p_one_shot) particles->one_shot = p_one_shot; } -void RendererStorageRD::particles_set_pre_process_time(RID p_particles, float p_time) { +void RendererStorageRD::particles_set_pre_process_time(RID p_particles, double p_time) { Particles *particles = particles_owner.getornull(p_particles); ERR_FAIL_COND(!particles); particles->pre_process_time = p_time; } -void RendererStorageRD::particles_set_explosiveness_ratio(RID p_particles, float p_ratio) { +void RendererStorageRD::particles_set_explosiveness_ratio(RID p_particles, real_t p_ratio) { Particles *particles = particles_owner.getornull(p_particles); ERR_FAIL_COND(!particles); particles->explosiveness = p_ratio; } -void RendererStorageRD::particles_set_randomness_ratio(RID p_particles, float p_ratio) { +void RendererStorageRD::particles_set_randomness_ratio(RID p_particles, real_t p_ratio) { Particles *particles = particles_owner.getornull(p_particles); ERR_FAIL_COND(!particles); particles->randomness = p_ratio; @@ -4003,7 +4127,7 @@ void RendererStorageRD::particles_set_custom_aabb(RID p_particles, const AABB &p particles->dependency.changed_notify(DEPENDENCY_CHANGED_AABB); } -void RendererStorageRD::particles_set_speed_scale(RID p_particles, float p_scale) { +void RendererStorageRD::particles_set_speed_scale(RID p_particles, double p_scale) { Particles *particles = particles_owner.getornull(p_particles); ERR_FAIL_COND(!particles); @@ -4046,7 +4170,7 @@ void RendererStorageRD::particles_set_fractional_delta(RID p_particles, bool p_e particles->fractional_delta = p_enable; } -void RendererStorageRD::particles_set_trails(RID p_particles, bool p_enable, float p_length) { +void RendererStorageRD::particles_set_trails(RID p_particles, bool p_enable, double p_length) { Particles *particles = particles_owner.getornull(p_particles); ERR_FAIL_COND(!particles); ERR_FAIL_COND(p_length < 0.1); @@ -4082,7 +4206,7 @@ void RendererStorageRD::particles_set_trail_bind_poses(RID p_particles, const Ve particles->dependency.changed_notify(DEPENDENCY_CHANGED_PARTICLES); } -void RendererStorageRD::particles_set_collision_base_size(RID p_particles, float p_size) { +void RendererStorageRD::particles_set_collision_base_size(RID p_particles, real_t p_size) { Particles *particles = particles_owner.getornull(p_particles); ERR_FAIL_COND(!particles); @@ -4216,6 +4340,10 @@ void RendererStorageRD::particles_request_process(RID p_particles) { } AABB RendererStorageRD::particles_get_current_aabb(RID p_particles) { + if (RSG::threaded) { + WARN_PRINT_ONCE("Calling this function with threaded rendering enabled stalls the renderer, use with care."); + } + const Particles *particles = particles_owner.getornull(p_particles); ERR_FAIL_COND_V(!particles, AABB()); @@ -4315,7 +4443,7 @@ void RendererStorageRD::particles_set_canvas_sdf_collision(RID p_particles, bool particles->sdf_collision_texture = p_texture; } -void RendererStorageRD::_particles_process(Particles *p_particles, float p_delta) { +void RendererStorageRD::_particles_process(Particles *p_particles, double p_delta) { if (p_particles->particles_material_uniform_set.is_null() || !RD::get_singleton()->uniform_set_is_valid(p_particles->particles_material_uniform_set)) { Vector<RD::Uniform> uniforms; @@ -4364,7 +4492,7 @@ void RendererStorageRD::_particles_process(Particles *p_particles, float p_delta p_particles->particles_material_uniform_set = RD::get_singleton()->uniform_set_create(uniforms, particles_shader.default_shader_rd, 1); } - float new_phase = Math::fmod((float)p_particles->phase + (p_delta / p_particles->lifetime) * p_particles->speed_scale, (float)1.0); + double new_phase = Math::fmod((double)p_particles->phase + (p_delta / p_particles->lifetime) * p_particles->speed_scale, 1.0); //move back history (if there is any) for (uint32_t i = p_particles->frame_history.size() - 1; i > 0; i--) { @@ -4720,7 +4848,7 @@ void RendererStorageRD::_particles_process(Particles *p_particles, float p_delta RD::get_singleton()->compute_list_bind_uniform_set(compute_list, p_particles->particles_material_uniform_set, 1); RD::get_singleton()->compute_list_bind_uniform_set(compute_list, p_particles->collision_textures_uniform_set, 2); - if (m->uniform_set.is_valid()) { + if (m->uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(m->uniform_set)) { RD::get_singleton()->compute_list_bind_uniform_set(compute_list, m->uniform_set, 3); } @@ -4832,7 +4960,7 @@ void RendererStorageRD::particles_set_view_axis(RID p_particles, const Vector3 & RD::get_singleton()->compute_list_dispatch_threads(compute_list, particles->amount, 1, 1); RD::get_singleton()->compute_list_end(); - effects.sort_buffer(particles->particles_sort_uniform_set, particles->amount); + effects->sort_buffer(particles->particles_sort_uniform_set, particles->amount); } copy_push_constant.total_particles *= copy_push_constant.total_particles; @@ -5004,14 +5132,14 @@ void RendererStorageRD::update_particles() { bool zero_time_scale = Engine::get_singleton()->get_time_scale() <= 0.0; if (particles->clear && particles->pre_process_time > 0.0) { - float frame_time; + double frame_time; if (fixed_fps > 0) { frame_time = 1.0 / fixed_fps; } else { frame_time = 1.0 / 30.0; } - float todo = particles->pre_process_time; + double todo = particles->pre_process_time; while (todo >= 0) { _particles_process(particles, frame_time); @@ -5020,8 +5148,8 @@ void RendererStorageRD::update_particles() { } if (fixed_fps > 0) { - float frame_time; - float decr; + double frame_time; + double decr; if (zero_time_scale) { frame_time = 0.0; decr = 1.0 / fixed_fps; @@ -5029,13 +5157,13 @@ void RendererStorageRD::update_particles() { frame_time = 1.0 / fixed_fps; decr = frame_time; } - float delta = RendererCompositorRD::singleton->get_frame_delta_time(); + double delta = RendererCompositorRD::singleton->get_frame_delta_time(); if (delta > 0.1) { //avoid recursive stalls if fps goes below 10 delta = 0.1; } else if (delta <= 0.0) { //unlikely but.. delta = 0.001; } - float todo = particles->frame_remainder + delta; + double todo = particles->frame_remainder + delta; while (todo >= frame_time) { _particles_process(particles, frame_time); @@ -5100,6 +5228,7 @@ void RendererStorageRD::update_particles() { } bool RendererStorageRD::particles_is_inactive(RID p_particles) const { + ERR_FAIL_COND_V_MSG(RSG::threaded, false, "This function should never be used with threaded rendering, as it stalls the renderer."); const Particles *particles = particles_owner.getornull(p_particles); ERR_FAIL_COND_V(!particles, false); return !particles->emitting && particles->inactive; @@ -5139,8 +5268,7 @@ void RendererStorageRD::ParticlesShaderData::set_code(const String &p_code) { actions.uniforms = &uniforms; Error err = base_singleton->particles_shader.compiler.compile(RS::SHADER_PARTICLES, code, &actions, path, gen_code); - - ERR_FAIL_COND(err != OK); + ERR_FAIL_COND_MSG(err != OK, "Shader compilation failed."); if (version.is_null()) { version = base_singleton->particles_shader.shader.version_create(); @@ -5250,94 +5378,14 @@ RendererStorageRD::ShaderData *RendererStorageRD::_create_particles_shader_func( return shader_data; } -void RendererStorageRD::ParticlesMaterialData::update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) { +bool RendererStorageRD::ParticlesMaterialData::update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) { uniform_set_updated = true; - if ((uint32_t)ubo_data.size() != shader_data->ubo_size) { - p_uniform_dirty = true; - if (uniform_buffer.is_valid()) { - RD::get_singleton()->free(uniform_buffer); - uniform_buffer = RID(); - } - - ubo_data.resize(shader_data->ubo_size); - if (ubo_data.size()) { - uniform_buffer = RD::get_singleton()->uniform_buffer_create(ubo_data.size()); - memset(ubo_data.ptrw(), 0, ubo_data.size()); //clear - } - - //clear previous uniform set - if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - RD::get_singleton()->free(uniform_set); - uniform_set = RID(); - } - } - - //check whether buffer changed - if (p_uniform_dirty && ubo_data.size()) { - update_uniform_buffer(shader_data->uniforms, shader_data->ubo_offsets.ptr(), p_parameters, ubo_data.ptrw(), ubo_data.size(), false); - RD::get_singleton()->buffer_update(uniform_buffer, 0, ubo_data.size(), ubo_data.ptrw()); - } - - uint32_t tex_uniform_count = shader_data->texture_uniforms.size(); - - if ((uint32_t)texture_cache.size() != tex_uniform_count) { - texture_cache.resize(tex_uniform_count); - p_textures_dirty = true; - - //clear previous uniform set - if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - RD::get_singleton()->free(uniform_set); - uniform_set = RID(); - } - } - - if (p_textures_dirty && tex_uniform_count) { - update_textures(p_parameters, shader_data->default_texture_params, shader_data->texture_uniforms, texture_cache.ptrw(), true); - } - - if (shader_data->ubo_size == 0 && shader_data->texture_uniforms.size() == 0) { - // This material does not require an uniform set, so don't create it. - return; - } - - if (!p_textures_dirty && uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - //no reason to update uniform set, only UBO (or nothing) was needed to update - return; - } - - Vector<RD::Uniform> uniforms; - - { - if (shader_data->ubo_size) { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_UNIFORM_BUFFER; - u.binding = 0; - u.ids.push_back(uniform_buffer); - uniforms.push_back(u); - } - - const RID *textures = texture_cache.ptrw(); - for (uint32_t i = 0; i < tex_uniform_count; i++) { - RD::Uniform u; - u.uniform_type = RD::UNIFORM_TYPE_TEXTURE; - u.binding = 1 + i; - u.ids.push_back(textures[i]); - uniforms.push_back(u); - } - } - - uniform_set = RD::get_singleton()->uniform_set_create(uniforms, base_singleton->particles_shader.shader.version_get_shader(shader_data->version, 0), 3); + return update_parameters_uniform_set(p_parameters, p_uniform_dirty, p_textures_dirty, shader_data->uniforms, shader_data->ubo_offsets.ptr(), shader_data->texture_uniforms, shader_data->default_texture_params, shader_data->ubo_size, uniform_set, base_singleton->particles_shader.shader.version_get_shader(shader_data->version, 0), 3); } RendererStorageRD::ParticlesMaterialData::~ParticlesMaterialData() { - if (uniform_set.is_valid() && RD::get_singleton()->uniform_set_is_valid(uniform_set)) { - RD::get_singleton()->free(uniform_set); - } - - if (uniform_buffer.is_valid()) { - RD::get_singleton()->free(uniform_buffer); - } + free_parameters_uniform_set(uniform_set); } RendererStorageRD::MaterialData *RendererStorageRD::_create_particles_material_func(ParticlesShaderData *p_shader) { @@ -5415,7 +5463,7 @@ void RendererStorageRD::particles_collision_set_cull_mask(RID p_particles_collis particles_collision->cull_mask = p_cull_mask; } -void RendererStorageRD::particles_collision_set_sphere_radius(RID p_particles_collision, float p_radius) { +void RendererStorageRD::particles_collision_set_sphere_radius(RID p_particles_collision, real_t p_radius) { ParticlesCollision *particles_collision = particles_collision_owner.getornull(p_particles_collision); ERR_FAIL_COND(!particles_collision); @@ -5431,21 +5479,21 @@ void RendererStorageRD::particles_collision_set_box_extents(RID p_particles_coll particles_collision->dependency.changed_notify(DEPENDENCY_CHANGED_AABB); } -void RendererStorageRD::particles_collision_set_attractor_strength(RID p_particles_collision, float p_strength) { +void RendererStorageRD::particles_collision_set_attractor_strength(RID p_particles_collision, real_t p_strength) { ParticlesCollision *particles_collision = particles_collision_owner.getornull(p_particles_collision); ERR_FAIL_COND(!particles_collision); particles_collision->attractor_strength = p_strength; } -void RendererStorageRD::particles_collision_set_attractor_directionality(RID p_particles_collision, float p_directionality) { +void RendererStorageRD::particles_collision_set_attractor_directionality(RID p_particles_collision, real_t p_directionality) { ParticlesCollision *particles_collision = particles_collision_owner.getornull(p_particles_collision); ERR_FAIL_COND(!particles_collision); particles_collision->attractor_directionality = p_directionality; } -void RendererStorageRD::particles_collision_set_attractor_attenuation(RID p_particles_collision, float p_curve) { +void RendererStorageRD::particles_collision_set_attractor_attenuation(RID p_particles_collision, real_t p_curve) { ParticlesCollision *particles_collision = particles_collision_owner.getornull(p_particles_collision); ERR_FAIL_COND(!particles_collision); @@ -5533,6 +5581,59 @@ void RendererStorageRD::particles_collision_instance_set_active(RID p_collision_ pci->active = p_active; } +/* VISIBILITY NOTIFIER */ + +RID RendererStorageRD::visibility_notifier_allocate() { + return visibility_notifier_owner.allocate_rid(); +} +void RendererStorageRD::visibility_notifier_initialize(RID p_notifier) { + visibility_notifier_owner.initialize_rid(p_notifier, VisibilityNotifier()); +} +void RendererStorageRD::visibility_notifier_set_aabb(RID p_notifier, const AABB &p_aabb) { + VisibilityNotifier *vn = visibility_notifier_owner.getornull(p_notifier); + ERR_FAIL_COND(!vn); + vn->aabb = p_aabb; + vn->dependency.changed_notify(DEPENDENCY_CHANGED_AABB); +} +void RendererStorageRD::visibility_notifier_set_callbacks(RID p_notifier, const Callable &p_enter_callbable, const Callable &p_exit_callable) { + VisibilityNotifier *vn = visibility_notifier_owner.getornull(p_notifier); + ERR_FAIL_COND(!vn); + vn->enter_callback = p_enter_callbable; + vn->exit_callback = p_exit_callable; +} + +AABB RendererStorageRD::visibility_notifier_get_aabb(RID p_notifier) const { + const VisibilityNotifier *vn = visibility_notifier_owner.getornull(p_notifier); + ERR_FAIL_COND_V(!vn, AABB()); + return vn->aabb; +} +void RendererStorageRD::visibility_notifier_call(RID p_notifier, bool p_enter, bool p_deferred) { + VisibilityNotifier *vn = visibility_notifier_owner.getornull(p_notifier); + ERR_FAIL_COND(!vn); + + if (p_enter) { + if (!vn->enter_callback.is_null()) { + if (p_deferred) { + vn->enter_callback.call_deferred(nullptr, 0); + } else { + Variant r; + Callable::CallError ce; + vn->enter_callback.call(nullptr, 0, r, ce); + } + } + } else { + if (!vn->exit_callback.is_null()) { + if (p_deferred) { + vn->exit_callback.call_deferred(nullptr, 0); + } else { + Variant r; + Callable::CallError ce; + vn->exit_callback.call(nullptr, 0, r, ce); + } + } + } +} + /* SKELETON API */ RID RendererStorageRD::skeleton_allocate() { @@ -5784,6 +5885,10 @@ void RendererStorageRD::light_set_param(RID p_light, RS::LightParam p_param, flo ERR_FAIL_COND(!light); ERR_FAIL_INDEX(p_param, RS::LIGHT_PARAM_MAX); + if (light->param[p_param] == p_value) { + return; + } + switch (p_param) { case RS::LIGHT_PARAM_RANGE: case RS::LIGHT_PARAM_SPOT_ANGLE: @@ -5797,6 +5902,12 @@ void RendererStorageRD::light_set_param(RID p_light, RS::LightParam p_param, flo light->version++; light->dependency.changed_notify(DEPENDENCY_CHANGED_LIGHT); } break; + case RS::LIGHT_PARAM_SIZE: { + if ((light->param[p_param] > CMP_EPSILON) != (p_value > CMP_EPSILON)) { + //changing from no size to size and the opposite + light->dependency.changed_notify(DEPENDENCY_CHANGED_LIGHT_SOFT_SHADOW_AND_PROJECTOR); + } + } break; default: { } } @@ -5833,8 +5944,11 @@ void RendererStorageRD::light_set_projector(RID p_light, RID p_texture) { light->projector = p_texture; - if (light->type != RS::LIGHT_DIRECTIONAL && light->projector.is_valid()) { - texture_add_to_decal_atlas(light->projector, light->type == RS::LIGHT_OMNI); + if (light->type != RS::LIGHT_DIRECTIONAL) { + if (light->projector.is_valid()) { + texture_add_to_decal_atlas(light->projector, light->type == RS::LIGHT_OMNI); + } + light->dependency.changed_notify(DEPENDENCY_CHANGED_LIGHT_SOFT_SHADOW_AND_PROJECTOR); } } @@ -5948,20 +6062,6 @@ RS::LightDirectionalShadowMode RendererStorageRD::light_directional_get_shadow_m return light->directional_shadow_mode; } -void RendererStorageRD::light_directional_set_shadow_depth_range_mode(RID p_light, RS::LightDirectionalShadowDepthRangeMode p_range_mode) { - Light *light = light_owner.getornull(p_light); - ERR_FAIL_COND(!light); - - light->directional_range_mode = p_range_mode; -} - -RS::LightDirectionalShadowDepthRangeMode RendererStorageRD::light_directional_get_shadow_depth_range_mode(RID p_light) const { - const Light *light = light_owner.getornull(p_light); - ERR_FAIL_COND_V(!light, RS::LIGHT_DIRECTIONAL_SHADOW_DEPTH_RANGE_STABLE); - - return light->directional_range_mode; -} - uint32_t RendererStorageRD::light_get_max_sdfgi_cascade(RID p_light) { const Light *light = light_owner.getornull(p_light); ERR_FAIL_COND_V(!light, 0); @@ -6555,32 +6655,6 @@ float RendererStorageRD::voxel_gi_get_energy(RID p_voxel_gi) const { return voxel_gi->energy; } -void RendererStorageRD::voxel_gi_set_ao(RID p_voxel_gi, float p_ao) { - VoxelGI *voxel_gi = voxel_gi_owner.getornull(p_voxel_gi); - ERR_FAIL_COND(!voxel_gi); - - voxel_gi->ao = p_ao; -} - -float RendererStorageRD::voxel_gi_get_ao(RID p_voxel_gi) const { - VoxelGI *voxel_gi = voxel_gi_owner.getornull(p_voxel_gi); - ERR_FAIL_COND_V(!voxel_gi, 0); - return voxel_gi->ao; -} - -void RendererStorageRD::voxel_gi_set_ao_size(RID p_voxel_gi, float p_strength) { - VoxelGI *voxel_gi = voxel_gi_owner.getornull(p_voxel_gi); - ERR_FAIL_COND(!voxel_gi); - - voxel_gi->ao_size = p_strength; -} - -float RendererStorageRD::voxel_gi_get_ao_size(RID p_voxel_gi) const { - VoxelGI *voxel_gi = voxel_gi_owner.getornull(p_voxel_gi); - ERR_FAIL_COND_V(!voxel_gi, 0); - return voxel_gi->ao_size; -} - void RendererStorageRD::voxel_gi_set_bias(RID p_voxel_gi, float p_bias) { VoxelGI *voxel_gi = voxel_gi_owner.getornull(p_voxel_gi); ERR_FAIL_COND(!voxel_gi); @@ -7461,7 +7535,7 @@ void RendererStorageRD::render_target_copy_to_back_buffer(RID p_render_target, c //single texture copy for backbuffer //RD::get_singleton()->texture_copy(rt->color, rt->backbuffer_mipmap0, Vector3(region.position.x, region.position.y, 0), Vector3(region.position.x, region.position.y, 0), Vector3(region.size.x, region.size.y, 1), 0, 0, 0, 0, true); - effects.copy_to_rect(rt->color, rt->backbuffer_mipmap0, region, false, false, false, true, true); + effects->copy_to_rect(rt->color, rt->backbuffer_mipmap0, region, false, false, false, true, true); if (!p_gen_mipmaps) { return; @@ -7477,7 +7551,7 @@ void RendererStorageRD::render_target_copy_to_back_buffer(RID p_render_target, c region.size.y = MAX(1, region.size.y >> 1); const RenderTarget::BackbufferMipmap &mm = rt->backbuffer_mipmaps[i]; - effects.gaussian_blur(prev_texture, mm.mipmap, mm.mipmap_copy, region, true); + effects->gaussian_blur(prev_texture, mm.mipmap, mm.mipmap_copy, region, true); prev_texture = mm.mipmap; } } @@ -7500,7 +7574,7 @@ void RendererStorageRD::render_target_clear_back_buffer(RID p_render_target, con } //single texture copy for backbuffer - effects.set_color(rt->backbuffer_mipmap0, p_color, region, true); + effects->set_color(rt->backbuffer_mipmap0, p_color, region, true); } void RendererStorageRD::render_target_gen_back_buffer_mipmaps(RID p_render_target, const Rect2i &p_region) { @@ -7530,7 +7604,7 @@ void RendererStorageRD::render_target_gen_back_buffer_mipmaps(RID p_render_targe region.size.y = MAX(1, region.size.y >> 1); const RenderTarget::BackbufferMipmap &mm = rt->backbuffer_mipmaps[i]; - effects.gaussian_blur(prev_texture, mm.mipmap, mm.mipmap_copy, region, true); + effects->gaussian_blur(prev_texture, mm.mipmap, mm.mipmap_copy, region, true); prev_texture = mm.mipmap; } } @@ -7588,6 +7662,9 @@ void RendererStorageRD::base_update_dependency(RID p_base, DependencyTracker *p_ } else if (particles_collision_owner.owns(p_base)) { ParticlesCollision *pc = particles_collision_owner.getornull(p_base); p_instance->update_dependency(&pc->dependency); + } else if (visibility_notifier_owner.owns(p_base)) { + VisibilityNotifier *vn = visibility_notifier_owner.getornull(p_base); + p_instance->update_dependency(&vn->dependency); } } @@ -7626,6 +7703,9 @@ RS::InstanceType RendererStorageRD::get_base_type(RID p_rid) const { if (particles_collision_owner.owns(p_rid)) { return RS::INSTANCE_PARTICLES_COLLISION; } + if (visibility_notifier_owner.owns(p_rid)) { + return RS::INSTANCE_VISIBLITY_NOTIFIER; + } return RS::INSTANCE_NONE; } @@ -7845,14 +7925,14 @@ void RendererStorageRD::_update_decal_atlas() { while ((K = decal_atlas.textures.next(K))) { DecalAtlas::Texture *t = decal_atlas.textures.getptr(*K); Texture *src_tex = texture_owner.getornull(*K); - effects.copy_to_atlas_fb(src_tex->rd_texture, mm.fb, t->uv_rect, draw_list, false, t->panorama_to_dp_users > 0); + effects->copy_to_atlas_fb(src_tex->rd_texture, mm.fb, t->uv_rect, draw_list, false, t->panorama_to_dp_users > 0); } RD::get_singleton()->draw_list_end(); prev_texture = mm.texture; } else { - effects.copy_to_fb_rect(prev_texture, mm.fb, Rect2i(Point2i(), mm.size)); + effects->copy_to_fb_rect(prev_texture, mm.fb, Rect2i(Point2i(), mm.size)); prev_texture = mm.texture; } } else { @@ -8272,6 +8352,9 @@ void RendererStorageRD::global_variable_set_override(const StringName &p_name, c if (!global_variables.variables.has(p_name)) { return; //variable may not exist } + + ERR_FAIL_COND(p_value.get_type() == Variant::OBJECT); + GlobalVariables::Variable &gv = global_variables.variables[p_name]; gv.override = p_value; @@ -8327,10 +8410,10 @@ void RendererStorageRD::global_variables_load_settings(bool p_load_textures) { List<PropertyInfo> settings; ProjectSettings::get_singleton()->get_property_list(&settings); - for (List<PropertyInfo>::Element *E = settings.front(); E; E = E->next()) { - if (E->get().name.begins_with("shader_globals/")) { - StringName name = E->get().name.get_slice("/", 1); - Dictionary d = ProjectSettings::get_singleton()->get(E->get().name); + for (const PropertyInfo &E : settings) { + if (E.name.begins_with("shader_globals/")) { + StringName name = E.name.get_slice("/", 1); + Dictionary d = ProjectSettings::get_singleton()->get(E.name); ERR_CONTINUE(!d.has("type")); ERR_CONTINUE(!d.has("value")); @@ -8498,8 +8581,8 @@ void RendererStorageRD::_update_global_variables() { if (global_variables.must_update_buffer_materials) { // only happens in the case of a buffer variable added or removed, // so not often. - for (List<RID>::Element *E = global_variables.materials_using_buffer.front(); E; E = E->next()) { - Material *material = material_owner.getornull(E->get()); + for (const RID &E : global_variables.materials_using_buffer) { + Material *material = material_owner.getornull(E); ERR_CONTINUE(!material); //wtf _material_queue_update(material, true, false); @@ -8511,8 +8594,8 @@ void RendererStorageRD::_update_global_variables() { if (global_variables.must_update_texture_materials) { // only happens in the case of a buffer variable added or removed, // so not often. - for (List<RID>::Element *E = global_variables.materials_using_texture.front(); E; E = E->next()) { - Material *material = material_owner.getornull(E->get()); + for (const RID &E : global_variables.materials_using_texture) { + Material *material = material_owner.getornull(E); ERR_CONTINUE(!material); //wtf _material_queue_update(material, false, true); @@ -8596,8 +8679,6 @@ bool RendererStorageRD::free(RID p_rid) { texture_owner.free(p_rid); } else if (canvas_texture_owner.owns(p_rid)) { - CanvasTexture *ct = canvas_texture_owner.getornull(p_rid); - memdelete(ct); canvas_texture_owner.free(p_rid); } else if (shader_owner.owns(p_rid)) { Shader *shader = shader_owner.getornull(p_rid); @@ -8613,9 +8694,6 @@ bool RendererStorageRD::free(RID p_rid) { } else if (material_owner.owns(p_rid)) { Material *material = material_owner.getornull(p_rid); - if (material->update_requested) { - _update_queued_materials(); - } material_set_shader(p_rid, RID()); //clean up shader material->dependency.deleted_notify(p_rid); @@ -8643,7 +8721,6 @@ bool RendererStorageRD::free(RID p_rid) { mi->I = nullptr; mesh_instance_owner.free(p_rid); - memdelete(mi); } else if (multimesh_owner.owns(p_rid)) { _update_dirty_multimeshes(); @@ -8702,6 +8779,10 @@ bool RendererStorageRD::free(RID p_rid) { } particles_collision->dependency.deleted_notify(p_rid); particles_collision_owner.free(p_rid); + } else if (visibility_notifier_owner.owns(p_rid)) { + VisibilityNotifier *vn = visibility_notifier_owner.getornull(p_rid); + vn->dependency.deleted_notify(p_rid); + visibility_notifier_owner.free(p_rid); } else if (particles_collision_instance_owner.owns(p_rid)) { particles_collision_instance_owner.free(p_rid); } else if (render_target_owner.owns(p_rid)) { @@ -8723,8 +8804,13 @@ bool RendererStorageRD::free(RID p_rid) { return true; } +void RendererStorageRD::init_effects(bool p_prefer_raster_effects) { + effects = memnew(EffectsRD(p_prefer_raster_effects)); +} + EffectsRD *RendererStorageRD::get_effects() { - return &effects; + ERR_FAIL_NULL_V_MSG(effects, nullptr, "Effects haven't been initialised yet."); + return effects; } void RendererStorageRD::capture_timestamps_begin() { @@ -8755,6 +8841,29 @@ String RendererStorageRD::get_captured_timestamp_name(uint32_t p_index) const { return RD::get_singleton()->get_captured_timestamp_name(p_index); } +void RendererStorageRD::update_memory_info() { + texture_mem_cache = RenderingDevice::get_singleton()->get_memory_usage(RenderingDevice::MEMORY_TEXTURES); + buffer_mem_cache = RenderingDevice::get_singleton()->get_memory_usage(RenderingDevice::MEMORY_BUFFERS); + total_mem_cache = RenderingDevice::get_singleton()->get_memory_usage(RenderingDevice::MEMORY_TOTAL); +} +uint64_t RendererStorageRD::get_rendering_info(RS::RenderingInfo p_info) { + if (p_info == RS::RENDERING_INFO_TEXTURE_MEM_USED) { + return texture_mem_cache; + } else if (p_info == RS::RENDERING_INFO_BUFFER_MEM_USED) { + return buffer_mem_cache; + } else if (p_info == RS::RENDERING_INFO_VIDEO_MEM_USED) { + return total_mem_cache; + } + return 0; +} + +String RendererStorageRD::get_video_adapter_name() const { + return RenderingDevice::get_singleton()->get_device_name(); +} +String RendererStorageRD::get_video_adapter_vendor() const { + return RenderingDevice::get_singleton()->get_device_vendor_name(); +} + RendererStorageRD *RendererStorageRD::base_singleton = nullptr; RendererStorageRD::RendererStorageRD() { @@ -8775,7 +8884,6 @@ RendererStorageRD::RendererStorageRD() { memset(global_variables.buffer_dirty_regions, 0, sizeof(bool) * global_variables.buffer_size / GlobalVariables::BUFFER_DIRTY_REGION_SIZE); global_variables.buffer = RD::get_singleton()->storage_buffer_create(sizeof(GlobalVariables::Value) * global_variables.buffer_size); - material_update_list = nullptr; { //create default textures RD::TextureFormat tformat; @@ -9020,26 +9128,42 @@ RendererStorageRD::RendererStorageRD() { } break; case RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS: { sampler_state.mag_filter = RD::SAMPLER_FILTER_NEAREST; - sampler_state.min_filter = RD::SAMPLER_FILTER_LINEAR; - sampler_state.mip_filter = RD::SAMPLER_FILTER_LINEAR; + sampler_state.min_filter = RD::SAMPLER_FILTER_NEAREST; + if (GLOBAL_GET("rendering/textures/default_filters/use_nearest_mipmap_filter")) { + sampler_state.mip_filter = RD::SAMPLER_FILTER_NEAREST; + } else { + sampler_state.mip_filter = RD::SAMPLER_FILTER_LINEAR; + } } break; case RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS: { sampler_state.mag_filter = RD::SAMPLER_FILTER_LINEAR; sampler_state.min_filter = RD::SAMPLER_FILTER_LINEAR; - sampler_state.mip_filter = RD::SAMPLER_FILTER_LINEAR; + if (GLOBAL_GET("rendering/textures/default_filters/use_nearest_mipmap_filter")) { + sampler_state.mip_filter = RD::SAMPLER_FILTER_NEAREST; + } else { + sampler_state.mip_filter = RD::SAMPLER_FILTER_LINEAR; + } } break; case RS::CANVAS_ITEM_TEXTURE_FILTER_NEAREST_WITH_MIPMAPS_ANISOTROPIC: { sampler_state.mag_filter = RD::SAMPLER_FILTER_NEAREST; - sampler_state.min_filter = RD::SAMPLER_FILTER_LINEAR; - sampler_state.mip_filter = RD::SAMPLER_FILTER_LINEAR; + sampler_state.min_filter = RD::SAMPLER_FILTER_NEAREST; + if (GLOBAL_GET("rendering/textures/default_filters/use_nearest_mipmap_filter")) { + sampler_state.mip_filter = RD::SAMPLER_FILTER_NEAREST; + } else { + sampler_state.mip_filter = RD::SAMPLER_FILTER_LINEAR; + } sampler_state.use_anisotropy = true; sampler_state.anisotropy_max = 1 << int(GLOBAL_GET("rendering/textures/default_filters/anisotropic_filtering_level")); } break; case RS::CANVAS_ITEM_TEXTURE_FILTER_LINEAR_WITH_MIPMAPS_ANISOTROPIC: { sampler_state.mag_filter = RD::SAMPLER_FILTER_LINEAR; sampler_state.min_filter = RD::SAMPLER_FILTER_LINEAR; - sampler_state.mip_filter = RD::SAMPLER_FILTER_LINEAR; + if (GLOBAL_GET("rendering/textures/default_filters/use_nearest_mipmap_filter")) { + sampler_state.mip_filter = RD::SAMPLER_FILTER_NEAREST; + } else { + sampler_state.mip_filter = RD::SAMPLER_FILTER_LINEAR; + } sampler_state.use_anisotropy = true; sampler_state.anisotropy_max = 1 << int(GLOBAL_GET("rendering/textures/default_filters/anisotropic_filtering_level")); @@ -9186,15 +9310,6 @@ RendererStorageRD::RendererStorageRD() { } } - { - Vector<String> sdf_versions; - sdf_versions.push_back(""); //one only - voxel_gi_sdf_shader.initialize(sdf_versions); - voxel_gi_sdf_shader_version = voxel_gi_sdf_shader.version_create(); - voxel_gi_sdf_shader_version_shader = voxel_gi_sdf_shader.version_get_shader(voxel_gi_sdf_shader_version, 0); - voxel_gi_sdf_shader_pipeline = RD::get_singleton()->compute_pipeline_create(voxel_gi_sdf_shader_version_shader); - } - using_lightmap_array = true; // high end if (using_lightmap_array) { uint32_t textures_per_stage = RD::get_singleton()->limit_get(RD::LIMIT_MAX_TEXTURES_PER_SHADER_STAGE); @@ -9282,7 +9397,15 @@ RendererStorageRD::RendererStorageRD() { // default material and shader for particles shader particles_shader.default_shader = shader_allocate(); shader_initialize(particles_shader.default_shader); - shader_set_code(particles_shader.default_shader, "shader_type particles; void process() { COLOR = vec4(1.0); } \n"); + shader_set_code(particles_shader.default_shader, R"( +// Default particles shader. + +shader_type particles; + +void process() { + COLOR = vec4(1.0); +} +)"); particles_shader.default_material = material_allocate(); material_initialize(particles_shader.default_material); material_set_shader(particles_shader.default_material, particles_shader.default_shader); @@ -9408,7 +9531,6 @@ RendererStorageRD::~RendererStorageRD() { RD::get_singleton()->free(mesh_default_rd_buffers[i]); } - voxel_gi_sdf_shader.version_free(voxel_gi_sdf_shader_version); particles_shader.copy_shader.version_free(particles_shader.copy_shader_version); rt_sdf.shader.version_free(rt_sdf.shader_version); @@ -9426,4 +9548,9 @@ RendererStorageRD::~RendererStorageRD() { if (decal_atlas.texture.is_valid()) { RD::get_singleton()->free(decal_atlas.texture); } + + if (effects) { + memdelete(effects); + effects = NULL; + } } diff --git a/servers/rendering/renderer_rd/renderer_storage_rd.h b/servers/rendering/renderer_rd/renderer_storage_rd.h index 28a1044705..02395a884f 100644 --- a/servers/rendering/renderer_rd/renderer_storage_rd.h +++ b/servers/rendering/renderer_rd/renderer_storage_rd.h @@ -155,9 +155,13 @@ public: virtual void set_render_priority(int p_priority) = 0; virtual void set_next_pass(RID p_pass) = 0; - virtual void update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) = 0; + virtual bool update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty) = 0; virtual ~MaterialData(); + //to be used internally by update_parameters, in the most common configuration of material parameters + bool update_parameters_uniform_set(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty, const Map<StringName, ShaderLanguage::ShaderNode::Uniform> &p_uniforms, const uint32_t *p_uniform_offsets, const Vector<ShaderCompilerRD::GeneratedCode::Texture> &p_texture_uniforms, const Map<StringName, RID> &p_default_texture_params, uint32_t p_ubo_size, RID &uniform_set, RID p_shader, uint32_t p_shader_uniform_set, uint32_t p_barrier = RD::BARRIER_MASK_ALL); + void free_parameters_uniform_set(RID p_uniform_set); + private: friend class RendererStorageRD; RID self; @@ -165,8 +169,14 @@ public: List<RID>::Element *global_texture_E = nullptr; uint64_t global_textures_pass = 0; Map<StringName, uint64_t> used_global_textures; + + //internally by update_parameters_uniform_set + Vector<uint8_t> ubo_data; + RID uniform_buffer; + Vector<RID> texture_cache; }; typedef MaterialData *(*MaterialDataRequestFunction)(ShaderData *); + static void _material_uniform_set_erased(const RID &p_set, void *p_material); enum DefaultRDTexture { DEFAULT_RD_TEXTURE_WHITE, @@ -221,7 +231,7 @@ private: ~CanvasTexture(); }; - RID_PtrOwner<CanvasTexture, true> canvas_texture_owner; + RID_Owner<CanvasTexture, true> canvas_texture_owner; /* TEXTURE API */ struct Texture { @@ -373,25 +383,28 @@ private: struct Material { RID self; - MaterialData *data; - Shader *shader; + MaterialData *data = nullptr; + Shader *shader = nullptr; //shortcut to shader data and type - ShaderType shader_type; + ShaderType shader_type = SHADER_TYPE_MAX; uint32_t shader_id = 0; - bool update_requested; - bool uniform_dirty; - bool texture_dirty; - Material *update_next; + bool uniform_dirty = false; + bool texture_dirty = false; Map<StringName, Variant> params; - int32_t priority; + int32_t priority = 0; RID next_pass; + SelfList<Material> update_element; + Dependency dependency; + + Material() : + update_element(this) {} }; MaterialDataRequestFunction material_data_request_func[SHADER_TYPE_MAX]; mutable RID_Owner<Material, true> material_owner; - Material *material_update_list; + SelfList<Material>::List material_update_list; void _material_queue_update(Material *material, bool p_uniform, bool p_texture); void _update_queued_materials(); @@ -434,6 +447,7 @@ private: struct LOD { float edge_length = 0.0; + uint32_t index_count = 0; RID index_buffer; RID index_array; }; @@ -513,7 +527,7 @@ private: void _mesh_instance_clear(MeshInstance *mi); void _mesh_instance_add_surface(MeshInstance *mi, Mesh *mesh, uint32_t p_surface); - mutable RID_PtrOwner<MeshInstance> mesh_instance_owner; + mutable RID_Owner<MeshInstance> mesh_instance_owner; SelfList<MeshInstance>::List dirty_mesh_instance_weights; SelfList<MeshInstance>::List dirty_mesh_instance_arrays; @@ -649,7 +663,7 @@ private: uint32_t type; uint32_t texture_index; //texture index for vector field - float scale; + real_t scale; uint32_t pad[2]; }; @@ -658,8 +672,8 @@ private: float prev_system_phase; uint32_t cycle; - float explosiveness; - float randomness; + real_t explosiveness; + real_t randomness; float time; float delta; @@ -701,14 +715,14 @@ private: struct Particles { RS::ParticlesMode mode = RS::PARTICLES_MODE_3D; bool inactive = true; - float inactive_time = 0.0; + double inactive_time = 0.0; bool emitting = false; bool one_shot = false; int amount = 0; - float lifetime = 1.0; - float pre_process_time = 0.0; - float explosiveness = 0.0; - float randomness = 0.0; + double lifetime = 1.0; + double pre_process_time = 0.0; + real_t explosiveness = 0.0; + real_t randomness = 0.0; bool restart_request = false; AABB custom_aabb = AABB(Vector3(-4, -4, -4), Vector3(8, 8, 8)); bool use_local_coords = true; @@ -752,20 +766,20 @@ private: RID sub_emitter; - float phase = 0.0; - float prev_phase = 0.0; + double phase = 0.0; + double prev_phase = 0.0; uint64_t prev_ticks = 0; uint32_t random_seed = 0; uint32_t cycle_number = 0; - float speed_scale = 1.0; + double speed_scale = 1.0; int fixed_fps = 30; bool interpolate = true; bool fractional_delta = false; - float frame_remainder = 0; - float collision_base_size = 0.01; + double frame_remainder = 0; + real_t collision_base_size = 0.01; bool clear = true; @@ -782,7 +796,7 @@ private: Dependency dependency; - float trail_length = 1.0; + double trail_length = 1.0; bool trails_enabled = false; LocalVector<ParticlesFrameParams> frame_history; LocalVector<ParticlesFrameParams> trail_params; @@ -791,7 +805,7 @@ private: } }; - void _particles_process(Particles *p_particles, float p_delta); + void _particles_process(Particles *p_particles, double p_delta); void _particles_allocate_emission_buffer(Particles *particles); void _particles_free_data(Particles *particles); void _particles_update_buffers(Particles *particles); @@ -894,17 +908,14 @@ private: } struct ParticlesMaterialData : public MaterialData { - uint64_t last_frame; - ParticlesShaderData *shader_data; - RID uniform_buffer; + uint64_t last_frame = 0; + ParticlesShaderData *shader_data = nullptr; RID uniform_set; - Vector<RID> texture_cache; - Vector<uint8_t> ubo_data; - bool uniform_set_updated; + bool uniform_set_updated = false; virtual void set_render_priority(int p_priority) {} virtual void set_next_pass(RID p_pass) {} - virtual void update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty); + virtual bool update_parameters(const Map<StringName, Variant> &p_parameters, bool p_uniform_dirty, bool p_textures_dirty); virtual ~ParticlesMaterialData(); }; @@ -947,6 +958,17 @@ private: mutable RID_Owner<ParticlesCollisionInstance> particles_collision_instance_owner; + /* visibility_notifier */ + + struct VisibilityNotifier { + AABB aabb; + Callable enter_callback; + Callable exit_callback; + Dependency dependency; + }; + + mutable RID_Owner<VisibilityNotifier> visibility_notifier_owner; + /* Skeleton */ struct Skeleton { @@ -991,7 +1013,6 @@ private: uint32_t cull_mask = 0xFFFFFFFF; RS::LightOmniShadowMode omni_shadow_mode = RS::LIGHT_OMNI_SHADOW_DUAL_PARABOLOID; RS::LightDirectionalShadowMode directional_shadow_mode = RS::LIGHT_DIRECTIONAL_SHADOW_ORTHOGONAL; - RS::LightDirectionalShadowDepthRangeMode directional_range_mode = RS::LIGHT_DIRECTIONAL_SHADOW_DEPTH_RANGE_STABLE; bool directional_blend_splits = false; bool directional_sky_only = false; uint64_t version = 0; @@ -1065,8 +1086,6 @@ private: float dynamic_range = 4.0; float energy = 1.0; - float ao = 0.0; - float ao_size = 0.5; float bias = 1.4; float normal_bias = 0.0; float propagation = 0.7; @@ -1081,11 +1100,6 @@ private: Dependency dependency; }; - VoxelGiSdfShaderRD voxel_gi_sdf_shader; - RID voxel_gi_sdf_shader_version; - RID voxel_gi_sdf_shader_version_shader; - RID voxel_gi_sdf_shader_pipeline; - mutable RID_Owner<VoxelGI, true> voxel_gi_owner; /* REFLECTION PROBE */ @@ -1271,7 +1285,7 @@ private: void _update_global_variables(); /* EFFECTS */ - EffectsRD effects; + EffectsRD *effects = NULL; public: virtual bool can_create_resources_async() const; @@ -1287,7 +1301,6 @@ public: virtual void _texture_2d_update(RID p_texture, const Ref<Image> &p_image, int p_layer, bool p_immediate); - virtual void texture_2d_update_immediate(RID p_texture, const Ref<Image> &p_image, int p_layer = 0); //mostly used for video and streaming virtual void texture_2d_update(RID p_texture, const Ref<Image> &p_image, int p_layer = 0); virtual void texture_3d_update(RID p_texture, const Vector<Ref<Image>> &p_data); virtual void texture_proxy_update(RID p_texture, RID p_proxy_to); @@ -1446,7 +1459,9 @@ public: virtual void mesh_set_blend_shape_mode(RID p_mesh, RS::BlendShapeMode p_mode); virtual RS::BlendShapeMode mesh_get_blend_shape_mode(RID p_mesh) const; - virtual void mesh_surface_update_region(RID p_mesh, int p_surface, int p_offset, const Vector<uint8_t> &p_data); + virtual void mesh_surface_update_vertex_region(RID p_mesh, int p_surface, int p_offset, const Vector<uint8_t> &p_data); + virtual void mesh_surface_update_attribute_region(RID p_mesh, int p_surface, int p_offset, const Vector<uint8_t> &p_data); + virtual void mesh_surface_update_skin_region(RID p_mesh, int p_surface, int p_offset, const Vector<uint8_t> &p_data); virtual void mesh_surface_set_material(RID p_mesh, int p_surface, RID p_material); virtual RID mesh_surface_get_material(RID p_mesh, int p_surface) const; @@ -1515,10 +1530,18 @@ public: return s->lod_count > 0; } - _FORCE_INLINE_ uint32_t mesh_surface_get_lod(void *p_surface, float p_model_scale, float p_distance_threshold, float p_lod_threshold) const { + _FORCE_INLINE_ uint32_t mesh_surface_get_vertices_drawn_count(void *p_surface) const { + Mesh::Surface *s = reinterpret_cast<Mesh::Surface *>(p_surface); + return s->index_count ? s->index_count : s->vertex_count; + } + + _FORCE_INLINE_ uint32_t mesh_surface_get_lod(void *p_surface, float p_model_scale, float p_distance_threshold, float p_lod_threshold, uint32_t *r_index_count = nullptr) const { Mesh::Surface *s = reinterpret_cast<Mesh::Surface *>(p_surface); int32_t current_lod = -1; + if (r_index_count) { + *r_index_count = s->index_count; + } for (uint32_t i = 0; i < s->lod_count; i++) { float screen_size = s->lods[i].edge_length * p_model_scale / p_distance_threshold; if (screen_size > p_lod_threshold) { @@ -1529,6 +1552,9 @@ public: if (current_lod == -1) { return 0; } else { + if (r_index_count) { + *r_index_count = s->lods[current_lod].index_count; + } return current_lod + 1; } } @@ -1735,24 +1761,6 @@ public: return multimesh->uniform_set_2d; } - /* IMMEDIATE API */ - - RID immediate_allocate() { return RID(); } - void immediate_initialize(RID p_immediate) {} - - virtual void immediate_begin(RID p_immediate, RS::PrimitiveType p_rimitive, RID p_texture = RID()) {} - virtual void immediate_vertex(RID p_immediate, const Vector3 &p_vertex) {} - virtual void immediate_normal(RID p_immediate, const Vector3 &p_normal) {} - virtual void immediate_tangent(RID p_immediate, const Plane &p_tangent) {} - virtual void immediate_color(RID p_immediate, const Color &p_color) {} - virtual void immediate_uv(RID p_immediate, const Vector2 &tex_uv) {} - virtual void immediate_uv2(RID p_immediate, const Vector2 &tex_uv) {} - virtual void immediate_end(RID p_immediate) {} - virtual void immediate_clear(RID p_immediate) {} - virtual void immediate_set_material(RID p_immediate, RID p_material) {} - virtual RID immediate_get_material(RID p_immediate) const { return RID(); } - virtual AABB immediate_get_aabb(RID p_immediate) const { return AABB(); } - /* SKELETON API */ RID skeleton_allocate(); @@ -1821,8 +1829,6 @@ public: bool light_directional_get_blend_splits(RID p_light) const; void light_directional_set_sky_only(RID p_light, bool p_sky_only); bool light_directional_is_sky_only(RID p_light) const; - void light_directional_set_shadow_depth_range_mode(RID p_light, RS::LightDirectionalShadowDepthRangeMode p_range_mode); - RS::LightDirectionalShadowDepthRangeMode light_directional_get_shadow_depth_range_mode(RID p_light) const; RS::LightDirectionalShadowMode light_directional_get_shadow_mode(RID p_light); RS::LightOmniShadowMode light_omni_get_shadow_mode(RID p_light); @@ -1877,6 +1883,13 @@ public: return light->shadow; } + _FORCE_INLINE_ bool light_has_projector(RID p_light) const { + const Light *light = light_owner.getornull(p_light); + ERR_FAIL_COND_V(!light, RS::LIGHT_DIRECTIONAL); + + return texture_owner.owns(light->projector); + } + _FORCE_INLINE_ bool light_is_negative(RID p_light) const { const Light *light = light_owner.getornull(p_light); ERR_FAIL_COND_V(!light, RS::LIGHT_DIRECTIONAL); @@ -2045,12 +2058,6 @@ public: void voxel_gi_set_energy(RID p_voxel_gi, float p_energy); float voxel_gi_get_energy(RID p_voxel_gi) const; - void voxel_gi_set_ao(RID p_voxel_gi, float p_ao); - float voxel_gi_get_ao(RID p_voxel_gi) const; - - void voxel_gi_set_ao_size(RID p_voxel_gi, float p_strength); - float voxel_gi_get_ao_size(RID p_voxel_gi) const; - void voxel_gi_set_bias(RID p_voxel_gi, float p_bias); float voxel_gi_get_bias(RID p_voxel_gi) const; @@ -2132,22 +2139,22 @@ public: void particles_set_mode(RID p_particles, RS::ParticlesMode p_mode); void particles_set_emitting(RID p_particles, bool p_emitting); void particles_set_amount(RID p_particles, int p_amount); - void particles_set_lifetime(RID p_particles, float p_lifetime); + void particles_set_lifetime(RID p_particles, double p_lifetime); void particles_set_one_shot(RID p_particles, bool p_one_shot); - void particles_set_pre_process_time(RID p_particles, float p_time); - void particles_set_explosiveness_ratio(RID p_particles, float p_ratio); - void particles_set_randomness_ratio(RID p_particles, float p_ratio); + void particles_set_pre_process_time(RID p_particles, double p_time); + void particles_set_explosiveness_ratio(RID p_particles, real_t p_ratio); + void particles_set_randomness_ratio(RID p_particles, real_t p_ratio); void particles_set_custom_aabb(RID p_particles, const AABB &p_aabb); - void particles_set_speed_scale(RID p_particles, float p_scale); + void particles_set_speed_scale(RID p_particles, double p_scale); void particles_set_use_local_coordinates(RID p_particles, bool p_enable); void particles_set_process_material(RID p_particles, RID p_material); void particles_set_fixed_fps(RID p_particles, int p_fps); void particles_set_interpolate(RID p_particles, bool p_enable); void particles_set_fractional_delta(RID p_particles, bool p_enable); - void particles_set_collision_base_size(RID p_particles, float p_size); + void particles_set_collision_base_size(RID p_particles, real_t p_size); void particles_set_transform_align(RID p_particles, RS::ParticlesTransformAlign p_transform_align); - void particles_set_trails(RID p_particles, bool p_enable, float p_length); + void particles_set_trails(RID p_particles, bool p_enable, double p_length); void particles_set_trail_bind_poses(RID p_particles, const Vector<Transform3D> &p_bind_poses); void particles_restart(RID p_particles); @@ -2240,11 +2247,11 @@ public: virtual void particles_collision_set_collision_type(RID p_particles_collision, RS::ParticlesCollisionType p_type); virtual void particles_collision_set_cull_mask(RID p_particles_collision, uint32_t p_cull_mask); - virtual void particles_collision_set_sphere_radius(RID p_particles_collision, float p_radius); //for spheres + virtual void particles_collision_set_sphere_radius(RID p_particles_collision, real_t p_radius); //for spheres virtual void particles_collision_set_box_extents(RID p_particles_collision, const Vector3 &p_extents); //for non-spheres - virtual void particles_collision_set_attractor_strength(RID p_particles_collision, float p_strength); - virtual void particles_collision_set_attractor_directionality(RID p_particles_collision, float p_directionality); - virtual void particles_collision_set_attractor_attenuation(RID p_particles_collision, float p_curve); + virtual void particles_collision_set_attractor_strength(RID p_particles_collision, real_t p_strength); + virtual void particles_collision_set_attractor_directionality(RID p_particles_collision, real_t p_directionality); + virtual void particles_collision_set_attractor_attenuation(RID p_particles_collision, real_t p_curve); virtual void particles_collision_set_field_texture(RID p_particles_collision, RID p_texture); //for SDF and vector field, heightfield is dynamic virtual void particles_collision_height_field_update(RID p_particles_collision); //for SDF and vector field virtual void particles_collision_set_height_field_resolution(RID p_particles_collision, RS::ParticlesCollisionHeightfieldResolution p_resolution); //for SDF and vector field @@ -2253,6 +2260,14 @@ public: virtual bool particles_collision_is_heightfield(RID p_particles_collision) const; RID particles_collision_get_heightfield_framebuffer(RID p_particles_collision) const; + virtual RID visibility_notifier_allocate(); + virtual void visibility_notifier_initialize(RID p_notifier); + virtual void visibility_notifier_set_aabb(RID p_notifier, const AABB &p_aabb); + virtual void visibility_notifier_set_callbacks(RID p_notifier, const Callable &p_enter_callbable, const Callable &p_exit_callable); + + virtual AABB visibility_notifier_get_aabb(RID p_notifier) const; + virtual void visibility_notifier_call(RID p_notifier, bool p_enter, bool p_deferred); + //used from 2D and 3D virtual RID particles_collision_instance_create(RID p_collision); virtual void particles_collision_instance_set_transform(RID p_collision_instance, const Transform3D &p_transform); @@ -2331,13 +2346,16 @@ public: void set_debug_generate_wireframes(bool p_generate) {} - void render_info_begin_capture() {} - void render_info_end_capture() {} - int get_captured_render_info(RS::RenderInfo p_info) { return 0; } + //keep cached since it can be called form any thread + uint64_t texture_mem_cache = 0; + uint64_t buffer_mem_cache = 0; + uint64_t total_mem_cache = 0; + + virtual void update_memory_info(); + virtual uint64_t get_rendering_info(RS::RenderingInfo p_info); - uint64_t get_render_info(RS::RenderInfo p_info) { return 0; } - String get_video_adapter_name() const { return String(); } - String get_video_adapter_vendor() const { return String(); } + String get_video_adapter_name() const; + String get_video_adapter_vendor() const; virtual void capture_timestamps_begin(); virtual void capture_timestamp(const String &p_name); @@ -2351,6 +2369,7 @@ public: static RendererStorageRD *base_singleton; + void init_effects(bool p_prefer_raster_effects); EffectsRD *get_effects(); RendererStorageRD(); diff --git a/servers/rendering/renderer_rd/shader_compiler_rd.cpp b/servers/rendering/renderer_rd/shader_compiler_rd.cpp index b347197289..b95d4b642c 100644 --- a/servers/rendering/renderer_rd/shader_compiler_rd.cpp +++ b/servers/rendering/renderer_rd/shader_compiler_rd.cpp @@ -213,7 +213,7 @@ static String _interpstr(SL::DataInterpolation p_interp) { return ""; } -static String _prestr(SL::DataPrecision p_pres) { +static String _prestr(SL::DataPrecision p_pres, bool p_force_highp = false) { switch (p_pres) { case SL::PRECISION_LOWP: return "lowp "; @@ -222,7 +222,7 @@ static String _prestr(SL::DataPrecision p_pres) { case SL::PRECISION_HIGHP: return "highp "; case SL::PRECISION_DEFAULT: - return ""; + return p_force_highp ? "highp " : ""; } return ""; } @@ -571,7 +571,7 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge max_texture_uniforms++; } else { if (E->get().scope == SL::ShaderNode::Uniform::SCOPE_INSTANCE) { - continue; //instances are indexed directly, dont need index uniforms + continue; // Instances are indexed directly, don't need index uniforms. } max_uniforms++; @@ -605,7 +605,7 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge if (uniform.scope == SL::ShaderNode::Uniform::SCOPE_INSTANCE) { //insert, but don't generate any code. p_actions.uniforms->insert(uniform_name, uniform); - continue; //instances are indexed directly, dont need index uniforms + continue; // Instances are indexed directly, don't need index uniforms. } if (SL::is_sampler_type(uniform.type)) { ucode = "layout(set = " + itos(actions.texture_layout_set) + ", binding = " + itos(actions.base_texture_binding_index + uniform.texture_order) + ") uniform "; @@ -617,7 +617,7 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge //this is an integer to index the global table ucode += _typestr(ShaderLanguage::TYPE_UINT); } else { - ucode += _prestr(uniform.precision); + ucode += _prestr(uniform.precision, ShaderLanguage::is_float_type(uniform.type)); ucode += _typestr(uniform.type); } @@ -742,7 +742,7 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge String vcode; String interp_mode = _interpstr(varying.interpolation); - vcode += _prestr(varying.precision); + vcode += _prestr(varying.precision, ShaderLanguage::is_float_type(varying.type)); vcode += _typestr(varying.type); vcode += " " + _mkid(varying_name); if (varying.array_size > 0) { @@ -760,11 +760,11 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge if (var_frag_to_light.size() > 0) { String gcode = "\n\nstruct {\n"; - for (List<Pair<StringName, SL::ShaderNode::Varying>>::Element *E = var_frag_to_light.front(); E; E = E->next()) { - gcode += "\t" + _prestr(E->get().second.precision) + _typestr(E->get().second.type) + " " + _mkid(E->get().first); - if (E->get().second.array_size > 0) { + for (const Pair<StringName, SL::ShaderNode::Varying> &E : var_frag_to_light) { + gcode += "\t" + _prestr(E.second.precision) + _typestr(E.second.type) + " " + _mkid(E.first); + if (E.second.array_size > 0) { gcode += "["; - gcode += itos(E->get().second.array_size); + gcode += itos(E.second.array_size); gcode += "]"; } gcode += ";\n"; @@ -777,7 +777,7 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge const SL::ShaderNode::Constant &cnode = pnode->vconstants[i]; String gcode; gcode += "const "; - gcode += _prestr(cnode.precision); + gcode += _prestr(cnode.precision, ShaderLanguage::is_float_type(cnode.type)); if (cnode.type == SL::TYPE_STRUCT) { gcode += _mkid(cnode.type_str); } else { @@ -887,7 +887,7 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge SL::VariableNode *vnode = (SL::VariableNode *)p_node; bool use_fragment_varying = false; - if (!(p_actions.entry_point_stages.has(current_func_name) && p_actions.entry_point_stages[current_func_name] == STAGE_VERTEX)) { + if (!vnode->is_local && !(p_actions.entry_point_stages.has(current_func_name) && p_actions.entry_point_stages[current_func_name] == STAGE_VERTEX)) { if (p_assigning) { if (shader->varyings.has(vnode->name)) { use_fragment_varying = true; @@ -1037,7 +1037,7 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge SL::ArrayNode *anode = (SL::ArrayNode *)p_node; bool use_fragment_varying = false; - if (!(p_actions.entry_point_stages.has(current_func_name) && p_actions.entry_point_stages[current_func_name] == STAGE_VERTEX)) { + if (!anode->is_local && !(p_actions.entry_point_stages.has(current_func_name) && p_actions.entry_point_stages[current_func_name] == STAGE_VERTEX)) { if (anode->assign_expression != nullptr && shader->varyings.has(anode->name)) { use_fragment_varying = true; } else { @@ -1165,6 +1165,7 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge SL::VariableNode *vnode = (SL::VariableNode *)onode->arguments[0]; bool is_texture_func = false; + bool is_screen_texture = false; if (onode->op == SL::OP_STRUCT) { code += _mkid(vnode->name); } else if (onode->op == SL::OP_CONSTRUCT) { @@ -1197,6 +1198,7 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge const SL::VariableNode *varnode = static_cast<const SL::VariableNode *>(onode->arguments[i]); StringName texture_uniform = varnode->name; + is_screen_texture = (texture_uniform == "SCREEN_TEXTURE"); String sampler_name; @@ -1236,6 +1238,9 @@ String ShaderCompilerRD::_dump_node_code(const SL::Node *p_node, int p_level, Ge } } code += ")"; + if (is_screen_texture && actions.apply_luminance_multiplier) { + code = "(" + code + " * vec4(vec3(sc_luminance_multiplier), 1.0))"; + } } break; case SL::OP_INDEX: { code += _dump_node_code(onode->arguments[0], p_level, r_gen_code, p_actions, p_default_actions, p_assigning); @@ -1351,7 +1356,13 @@ Error ShaderCompilerRD::compile(RS::ShaderMode p_mode, const String &p_code, Ide if (err != OK) { Vector<String> shader = p_code.split("\n"); for (int i = 0; i < shader.size(); i++) { - print_line(itos(i + 1) + " " + shader[i]); + if (i + 1 == parser.get_error_line()) { + // Mark the error line to be visible without having to look at + // the trace at the end. + print_line(vformat("E%4d-> %s", i + 1, shader[i])); + } else { + print_line(vformat("%5d | %s", i + 1, shader[i])); + } } _err_print_error(nullptr, p_path.utf8().get_data(), parser.get_error_line(), parser.get_error_text().utf8().get_data(), ERR_HANDLER_SHADER); @@ -1388,8 +1399,8 @@ void ShaderCompilerRD::initialize(DefaultIdentifierActions p_actions) { ShaderLanguage::get_builtin_funcs(&func_list); - for (List<String>::Element *E = func_list.front(); E; E = E->next()) { - internal_functions.insert(E->get()); + for (const String &E : func_list) { + internal_functions.insert(E); } texture_functions.insert("texture"); texture_functions.insert("textureProj"); diff --git a/servers/rendering/renderer_rd/shader_compiler_rd.h b/servers/rendering/renderer_rd/shader_compiler_rd.h index 2da127ffa3..0fe9047967 100644 --- a/servers/rendering/renderer_rd/shader_compiler_rd.h +++ b/servers/rendering/renderer_rd/shader_compiler_rd.h @@ -95,6 +95,7 @@ public: String global_buffer_array_variable; String instance_uniform_index_variable; uint32_t base_varying_index = 0; + bool apply_luminance_multiplier = false; }; private: diff --git a/servers/rendering/renderer_rd/shader_rd.cpp b/servers/rendering/renderer_rd/shader_rd.cpp index 27305cc938..82efa1318c 100644 --- a/servers/rendering/renderer_rd/shader_rd.cpp +++ b/servers/rendering/renderer_rd/shader_rd.cpp @@ -116,8 +116,10 @@ void ShaderRD::setup(const char *p_vertex_code, const char *p_fragment_code, con } StringBuilder tohash; - tohash.append("[VersionKey]"); - tohash.append(RenderingDevice::get_singleton()->shader_get_cache_key()); + tohash.append("[SpirvCacheKey]"); + tohash.append(RenderingDevice::get_singleton()->shader_get_spirv_cache_key()); + tohash.append("[BinaryCacheKey]"); + tohash.append(RenderingDevice::get_singleton()->shader_get_binary_cache_key()); tohash.append("[Vertex]"); tohash.append(p_vertex_code ? p_vertex_code : ""); tohash.append("[Fragment]"); @@ -144,12 +146,14 @@ void ShaderRD::_clear_version(Version *p_version) { //clear versions if they exist if (p_version->variants) { for (int i = 0; i < variant_defines.size(); i++) { - RD::get_singleton()->free(p_version->variants[i]); + if (variants_enabled[i]) { + RD::get_singleton()->free(p_version->variants[i]); + } } memdelete_arr(p_version->variants); - if (p_version->variant_stages) { - memdelete_arr(p_version->variant_stages); + if (p_version->variant_data) { + memdelete_arr(p_version->variant_data); } p_version->variants = nullptr; } @@ -203,7 +207,7 @@ void ShaderRD::_compile_variant(uint32_t p_variant, Version *p_version) { return; //variant is disabled, return } - Vector<RD::ShaderStageData> &stages = p_version->variant_stages[p_variant]; + Vector<RD::ShaderStageSPIRVData> stages; String error; String current_source; @@ -217,8 +221,8 @@ void ShaderRD::_compile_variant(uint32_t p_variant, Version *p_version) { _build_variant_code(builder, p_variant, p_version, stage_templates[STAGE_TYPE_VERTEX]); current_source = builder.as_string(); - RD::ShaderStageData stage; - stage.spir_v = RD::get_singleton()->shader_compile_from_source(RD::SHADER_STAGE_VERTEX, current_source, RD::SHADER_LANGUAGE_GLSL, &error); + RD::ShaderStageSPIRVData stage; + stage.spir_v = RD::get_singleton()->shader_compile_spirv_from_source(RD::SHADER_STAGE_VERTEX, current_source, RD::SHADER_LANGUAGE_GLSL, &error); if (stage.spir_v.size() == 0) { build_ok = false; } else { @@ -235,8 +239,8 @@ void ShaderRD::_compile_variant(uint32_t p_variant, Version *p_version) { _build_variant_code(builder, p_variant, p_version, stage_templates[STAGE_TYPE_FRAGMENT]); current_source = builder.as_string(); - RD::ShaderStageData stage; - stage.spir_v = RD::get_singleton()->shader_compile_from_source(RD::SHADER_STAGE_FRAGMENT, current_source, RD::SHADER_LANGUAGE_GLSL, &error); + RD::ShaderStageSPIRVData stage; + stage.spir_v = RD::get_singleton()->shader_compile_spirv_from_source(RD::SHADER_STAGE_FRAGMENT, current_source, RD::SHADER_LANGUAGE_GLSL, &error); if (stage.spir_v.size() == 0) { build_ok = false; } else { @@ -254,8 +258,8 @@ void ShaderRD::_compile_variant(uint32_t p_variant, Version *p_version) { current_source = builder.as_string(); - RD::ShaderStageData stage; - stage.spir_v = RD::get_singleton()->shader_compile_from_source(RD::SHADER_STAGE_COMPUTE, current_source, RD::SHADER_LANGUAGE_GLSL, &error); + RD::ShaderStageSPIRVData stage; + stage.spir_v = RD::get_singleton()->shader_compile_spirv_from_source(RD::SHADER_STAGE_COMPUTE, current_source, RD::SHADER_LANGUAGE_GLSL, &error); if (stage.spir_v.size() == 0) { build_ok = false; } else { @@ -275,10 +279,15 @@ void ShaderRD::_compile_variant(uint32_t p_variant, Version *p_version) { return; } - RID shader = RD::get_singleton()->shader_create(stages); + Vector<uint8_t> shader_data = RD::get_singleton()->shader_compile_binary_from_spirv(stages, name + ":" + itos(p_variant)); + + ERR_FAIL_COND(shader_data.size() == 0); + + RID shader = RD::get_singleton()->shader_create_from_bytecode(shader_data); { MutexLock lock(variant_set_mutex); p_version->variants[p_variant] = shader; + p_version->variant_data[p_variant] = shader_data; } } @@ -364,14 +373,12 @@ String ShaderRD::_version_get_sha1(Version *p_version) const { } static const char *shader_file_header = "GDSC"; -static const uint32_t cache_file_version = 1; +static const uint32_t cache_file_version = 2; bool ShaderRD::_load_from_cache(Version *p_version) { String sha1 = _version_get_sha1(p_version); String path = shader_cache_dir.plus_file(name).plus_file(base_sha256).plus_file(sha1) + ".cache"; - uint64_t time_from = OS::get_singleton()->get_ticks_usec(); - FileAccessRef f = FileAccess::open(path, FileAccess::READ); if (!f) { return false; @@ -390,76 +397,43 @@ bool ShaderRD::_load_from_cache(Version *p_version) { ERR_FAIL_COND_V(variant_count != (uint32_t)variant_defines.size(), false); //should not happen but check - bool success = true; for (uint32_t i = 0; i < variant_count; i++) { - uint32_t stage_count = f->get_32(); - p_version->variant_stages[i].resize(stage_count); - for (uint32_t j = 0; j < stage_count; j++) { - p_version->variant_stages[i].write[j].shader_stage = RD::ShaderStage(f->get_32()); - - int compression = f->get_32(); - uint32_t length = f->get_32(); - - if (compression == 0) { - Vector<uint8_t> data; - data.resize(length); - - f->get_buffer(data.ptrw(), length); - - p_version->variant_stages[i].write[j].spir_v = data; - } else { - Vector<uint8_t> data; - - if (compression == 2) { - //zstd - int smol_length = f->get_32(); - Vector<uint8_t> zstd_data; - - zstd_data.resize(smol_length); - f->get_buffer(zstd_data.ptrw(), smol_length); - - data.resize(length); - Compression::decompress(data.ptrw(), data.size(), zstd_data.ptr(), zstd_data.size(), Compression::MODE_ZSTD); - - } else { - data.resize(length); - f->get_buffer(data.ptrw(), length); - } - - Vector<uint8_t> spirv; - uint32_t spirv_size = smolv::GetDecodedBufferSize(data.ptr(), data.size()); - spirv.resize(spirv_size); - if (!smolv::Decode(data.ptr(), data.size(), spirv.ptrw(), spirv_size)) { - ERR_PRINT("Malformed smolv input uncompressing shader " + name + ", variant #" + itos(i) + " stage :" + itos(j)); - success = false; - break; - } - p_version->variant_stages[i].write[j].spir_v = spirv; - } + uint32_t variant_size = f->get_32(); + ERR_FAIL_COND_V(variant_size == 0 && variants_enabled[i], false); + if (!variants_enabled[i]) { + continue; } - } + Vector<uint8_t> variant_bytes; + variant_bytes.resize(variant_size); - if (!success) { - for (uint32_t i = 0; i < variant_count; i++) { - p_version->variant_stages[i].resize(0); - } - return false; - } + uint32_t br = f->get_buffer(variant_bytes.ptrw(), variant_size); - float time_ms = double(OS::get_singleton()->get_ticks_usec() - time_from) / 1000.0; + ERR_FAIL_COND_V(br != variant_size, false); - print_verbose("Shader cache load success '" + path + "' " + rtos(time_ms) + "ms."); + p_version->variant_data[i] = variant_bytes; + } for (uint32_t i = 0; i < variant_count; i++) { - RID shader = RD::get_singleton()->shader_create(p_version->variant_stages[i]); + if (!variants_enabled[i]) { + MutexLock lock(variant_set_mutex); + p_version->variants[i] = RID(); + continue; + } + RID shader = RD::get_singleton()->shader_create_from_bytecode(p_version->variant_data[i]); + if (shader.is_null()) { + for (uint32_t j = 0; j < i; j++) { + RD::get_singleton()->free(p_version->variants[i]); + } + ERR_FAIL_COND_V(shader.is_null(), false); + } { MutexLock lock(variant_set_mutex); p_version->variants[i] = shader; } } - memdelete_arr(p_version->variant_stages); //clear stages - p_version->variant_stages = nullptr; + memdelete_arr(p_version->variant_data); //clear stages + p_version->variant_data = nullptr; p_version->valid = true; return true; } @@ -476,49 +450,8 @@ void ShaderRD::_save_to_cache(Version *p_version) { f->store_32(variant_count); //variant count for (uint32_t i = 0; i < variant_count; i++) { - f->store_32(p_version->variant_stages[i].size()); //stage count - for (int j = 0; j < p_version->variant_stages[i].size(); j++) { - f->store_32(p_version->variant_stages[i][j].shader_stage); //stage count - Vector<uint8_t> spirv = p_version->variant_stages[i][j].spir_v; - - bool save_uncompressed = true; - if (shader_cache_save_compressed) { - smolv::ByteArray smolv; - bool strip_debug = !shader_cache_save_debug; - if (!smolv::Encode(spirv.ptr(), spirv.size(), smolv, strip_debug ? smolv::kEncodeFlagStripDebugInfo : 0)) { - ERR_PRINT("Error compressing shader " + name + ", variant #" + itos(i) + " stage :" + itos(i)); - } else { - bool compress_zstd = shader_cache_save_compressed_zstd; - - if (compress_zstd) { - Vector<uint8_t> zstd; - zstd.resize(Compression::get_max_compressed_buffer_size(smolv.size(), Compression::MODE_ZSTD)); - int dst_size = Compression::compress(zstd.ptrw(), &smolv[0], smolv.size(), Compression::MODE_ZSTD); - if (dst_size >= 0 && (uint32_t)dst_size < smolv.size()) { - f->store_32(2); //compressed zstd - f->store_32(smolv.size()); //size of smolv buffer - f->store_32(dst_size); //size of smolv buffer - f->store_buffer(zstd.ptr(), dst_size); //smolv buffer - } else { - compress_zstd = false; - } - } - - if (!compress_zstd) { - f->store_32(1); //compressed - f->store_32(smolv.size()); //size of smolv buffer - f->store_buffer(&smolv[0], smolv.size()); //smolv buffer - } - save_uncompressed = false; - } - } - - if (save_uncompressed) { - f->store_32(0); //uncompressed - f->store_32(spirv.size()); //stage count - f->store_buffer(spirv.ptr(), spirv.size()); //stage count - } - } + f->store_32(p_version->variant_data[i].size()); //stage count + f->store_buffer(p_version->variant_data[i].ptr(), p_version->variant_data[i].size()); } f->close(); @@ -531,8 +464,8 @@ void ShaderRD::_compile_version(Version *p_version) { p_version->dirty = false; p_version->variants = memnew_arr(RID, variant_defines.size()); - typedef Vector<RD::ShaderStageData> ShaderStageArray; - p_version->variant_stages = memnew_arr(ShaderStageArray, variant_defines.size()); + typedef Vector<uint8_t> ShaderStageData; + p_version->variant_data = memnew_arr(ShaderStageData, variant_defines.size()); if (shader_cache_dir_valid) { if (_load_from_cache(p_version)) { @@ -571,19 +504,19 @@ void ShaderRD::_compile_version(Version *p_version) { } } memdelete_arr(p_version->variants); - if (p_version->variant_stages) { - memdelete_arr(p_version->variant_stages); + if (p_version->variant_data) { + memdelete_arr(p_version->variant_data); } p_version->variants = nullptr; - p_version->variant_stages = nullptr; + p_version->variant_data = nullptr; return; } else if (shader_cache_dir_valid) { //save shader cache _save_to_cache(p_version); } - memdelete_arr(p_version->variant_stages); //clear stages - p_version->variant_stages = nullptr; + memdelete_arr(p_version->variant_data); //clear stages + p_version->variant_data = nullptr; p_version->valid = true; } diff --git a/servers/rendering/renderer_rd/shader_rd.h b/servers/rendering/renderer_rd/shader_rd.h index 9a68e02007..529328f0ed 100644 --- a/servers/rendering/renderer_rd/shader_rd.h +++ b/servers/rendering/renderer_rd/shader_rd.h @@ -59,7 +59,7 @@ class ShaderRD { Map<StringName, CharString> code_sections; Vector<CharString> custom_defines; - Vector<RD::ShaderStageData> *variant_stages = nullptr; + Vector<uint8_t> *variant_data = nullptr; RID *variants = nullptr; //same size as version defines bool valid; diff --git a/servers/rendering/renderer_rd/shaders/blit.glsl b/servers/rendering/renderer_rd/shaders/blit.glsl index 967da1e6e4..8051f96738 100644 --- a/servers/rendering/renderer_rd/shaders/blit.glsl +++ b/servers/rendering/renderer_rd/shaders/blit.glsl @@ -5,6 +5,7 @@ #VERSION_DEFINES layout(push_constant, binding = 0, std140) uniform Pos { + vec4 src_rect; vec4 dst_rect; vec2 eye_center; @@ -22,8 +23,8 @@ layout(location = 0) out vec2 uv; void main() { vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); - uv = base_arr[gl_VertexIndex]; - vec2 vtx = data.dst_rect.xy + uv * data.dst_rect.zw; + uv = data.src_rect.xy + base_arr[gl_VertexIndex] * data.src_rect.zw; + vec2 vtx = data.dst_rect.xy + base_arr[gl_VertexIndex] * data.dst_rect.zw; gl_Position = vec4(vtx * 2.0 - 1.0, 0.0, 1.0); } @@ -34,6 +35,7 @@ void main() { #VERSION_DEFINES layout(push_constant, binding = 0, std140) uniform Pos { + vec4 src_rect; vec4 dst_rect; vec2 eye_center; diff --git a/servers/rendering/renderer_rd/shaders/blur_raster.glsl b/servers/rendering/renderer_rd/shaders/blur_raster.glsl new file mode 100644 index 0000000000..f8b4e3f610 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/blur_raster.glsl @@ -0,0 +1,138 @@ +/* clang-format off */ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +#include "blur_raster_inc.glsl" + +layout(location = 0) out vec2 uv_interp; +/* clang-format on */ + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv_interp = base_arr[gl_VertexIndex]; + + gl_Position = vec4(uv_interp * 2.0 - 1.0, 0.0, 1.0); +} + +/* clang-format off */ +#[fragment] + +#version 450 + +#VERSION_DEFINES + +#include "blur_raster_inc.glsl" + +layout(location = 0) in vec2 uv_interp; +/* clang-format on */ + +layout(set = 0, binding = 0) uniform sampler2D source_color; + +#ifdef GLOW_USE_AUTO_EXPOSURE +layout(set = 1, binding = 0) uniform sampler2D source_auto_exposure; +#endif + +layout(location = 0) out vec4 frag_color; + +void main() { + // We do not apply our color scale for our mobile renderer here, we'll leave our colors at half brightness and apply scale in the tonemap raster. + +#ifdef MODE_MIPMAP + + vec2 pix_size = blur.pixel_size; + vec4 color = texture(source_color, uv_interp + vec2(-0.5, -0.5) * pix_size); + color += texture(source_color, uv_interp + vec2(0.5, -0.5) * pix_size); + color += texture(source_color, uv_interp + vec2(0.5, 0.5) * pix_size); + color += texture(source_color, uv_interp + vec2(-0.5, 0.5) * pix_size); + frag_color = color / 4.0; + +#endif + +#ifdef MODE_GAUSSIAN_BLUR + + //Simpler blur uses SIGMA2 for the gaussian kernel for a stronger effect + + if (bool(blur.flags & FLAG_HORIZONTAL)) { + vec2 pix_size = blur.pixel_size; + pix_size *= 0.5; //reading from larger buffer, so use more samples + vec4 color = texture(source_color, uv_interp + vec2(0.0, 0.0) * pix_size) * 0.214607; + color += texture(source_color, uv_interp + vec2(1.0, 0.0) * pix_size) * 0.189879; + color += texture(source_color, uv_interp + vec2(2.0, 0.0) * pix_size) * 0.131514; + color += texture(source_color, uv_interp + vec2(3.0, 0.0) * pix_size) * 0.071303; + color += texture(source_color, uv_interp + vec2(-1.0, 0.0) * pix_size) * 0.189879; + color += texture(source_color, uv_interp + vec2(-2.0, 0.0) * pix_size) * 0.131514; + color += texture(source_color, uv_interp + vec2(-3.0, 0.0) * pix_size) * 0.071303; + frag_color = color; + } else { + vec2 pix_size = blur.pixel_size; + vec4 color = texture(source_color, uv_interp + vec2(0.0, 0.0) * pix_size) * 0.38774; + color += texture(source_color, uv_interp + vec2(0.0, 1.0) * pix_size) * 0.24477; + color += texture(source_color, uv_interp + vec2(0.0, 2.0) * pix_size) * 0.06136; + color += texture(source_color, uv_interp + vec2(0.0, -1.0) * pix_size) * 0.24477; + color += texture(source_color, uv_interp + vec2(0.0, -2.0) * pix_size) * 0.06136; + frag_color = color; + } +#endif + +#ifdef MODE_GAUSSIAN_GLOW + + //Glow uses larger sigma 1 for a more rounded blur effect + +#define GLOW_ADD(m_ofs, m_mult) \ + { \ + vec2 ofs = uv_interp + m_ofs * pix_size; \ + vec4 c = texture(source_color, ofs) * m_mult; \ + if (any(lessThan(ofs, vec2(0.0))) || any(greaterThan(ofs, vec2(1.0)))) { \ + c *= 0.0; \ + } \ + color += c; \ + } + + if (bool(blur.flags & FLAG_HORIZONTAL)) { + vec2 pix_size = blur.pixel_size; + pix_size *= 0.5; //reading from larger buffer, so use more samples + vec4 color = texture(source_color, uv_interp + vec2(0.0, 0.0) * pix_size) * 0.174938; + GLOW_ADD(vec2(1.0, 0.0), 0.165569); + GLOW_ADD(vec2(2.0, 0.0), 0.140367); + GLOW_ADD(vec2(3.0, 0.0), 0.106595); + GLOW_ADD(vec2(-1.0, 0.0), 0.165569); + GLOW_ADD(vec2(-2.0, 0.0), 0.140367); + GLOW_ADD(vec2(-3.0, 0.0), 0.106595); + color *= blur.glow_strength; + frag_color = color; + } else { + vec2 pix_size = blur.pixel_size; + vec4 color = texture(source_color, uv_interp + vec2(0.0, 0.0) * pix_size) * 0.288713; + GLOW_ADD(vec2(0.0, 1.0), 0.233062); + GLOW_ADD(vec2(0.0, 2.0), 0.122581); + GLOW_ADD(vec2(0.0, -1.0), 0.233062); + GLOW_ADD(vec2(0.0, -2.0), 0.122581); + color *= blur.glow_strength; + frag_color = color; + } + +#undef GLOW_ADD + + if (bool(blur.flags & FLAG_GLOW_FIRST_PASS)) { +#ifdef GLOW_USE_AUTO_EXPOSURE + + frag_color /= texelFetch(source_auto_exposure, ivec2(0, 0), 0).r / blur.glow_auto_exposure_grey; +#endif + frag_color *= blur.glow_exposure; + + float luminance = max(frag_color.r, max(frag_color.g, frag_color.b)); + float feedback = max(smoothstep(blur.glow_hdr_threshold, blur.glow_hdr_threshold + blur.glow_hdr_scale, luminance), blur.glow_bloom); + + frag_color = min(frag_color * feedback, vec4(blur.glow_luminance_cap)); + } + +#endif + +#ifdef MODE_COPY + vec4 color = textureLod(source_color, uv_interp, 0.0); + frag_color = color; +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/blur_raster_inc.glsl b/servers/rendering/renderer_rd/shaders/blur_raster_inc.glsl new file mode 100644 index 0000000000..52bf2886b5 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/blur_raster_inc.glsl @@ -0,0 +1,21 @@ +#define FLAG_HORIZONTAL (1 << 0) +#define FLAG_USE_ORTHOGONAL_PROJECTION (1 << 1) +#define FLAG_GLOW_FIRST_PASS (1 << 2) + +layout(push_constant, binding = 1, std430) uniform Blur { + vec2 pixel_size; + uint flags; + uint pad; + + // Glow. + float glow_strength; + float glow_bloom; + float glow_hdr_threshold; + float glow_hdr_scale; + + float glow_exposure; + float glow_white; + float glow_luminance_cap; + float glow_auto_exposure_grey; +} +blur; diff --git a/servers/rendering/renderer_rd/shaders/bokeh_dof.glsl b/servers/rendering/renderer_rd/shaders/bokeh_dof.glsl index b70e0b6bd5..0438671dd2 100644 --- a/servers/rendering/renderer_rd/shaders/bokeh_dof.glsl +++ b/servers/rendering/renderer_rd/shaders/bokeh_dof.glsl @@ -25,34 +25,7 @@ layout(set = 1, binding = 0) uniform sampler2D source_bokeh; // based on https://www.shadertoy.com/view/Xd3GDl -layout(push_constant, binding = 1, std430) uniform Params { - ivec2 size; - float z_far; - float z_near; - - bool orthogonal; - float blur_size; - float blur_scale; - int blur_steps; - - bool blur_near_active; - float blur_near_begin; - float blur_near_end; - bool blur_far_active; - - float blur_far_begin; - float blur_far_end; - bool second_pass; - bool half_size; - - bool use_jitter; - float jitter_seed; - uint pad[2]; -} -params; - -//used to work around downsampling filter -#define DEPTH_GAP 0.0 +#include "bokeh_dof_inc.glsl" #ifdef MODE_GEN_BLUR_SIZE @@ -80,15 +53,6 @@ float get_blur_size(float depth) { #endif -const float GOLDEN_ANGLE = 2.39996323; - -//note: uniform pdf rand [0;1[ -float hash12n(vec2 p) { - p = fract(p * vec2(5.3987, 5.4421)); - p += dot(p.yx, p.xy + vec2(21.5351, 14.3137)); - return fract(p.x * p.y * 95.4307); -} - #if defined(MODE_BOKEH_BOX) || defined(MODE_BOKEH_HEXAGONAL) vec4 weighted_filter_dir(vec2 dir, vec2 uv, vec2 pixel_size) { diff --git a/servers/rendering/renderer_rd/shaders/bokeh_dof_inc.glsl b/servers/rendering/renderer_rd/shaders/bokeh_dof_inc.glsl new file mode 100644 index 0000000000..fadea1631c --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/bokeh_dof_inc.glsl @@ -0,0 +1,37 @@ +layout(push_constant, binding = 1, std430) uniform Params { + ivec2 size; + float z_far; + float z_near; + + bool orthogonal; + float blur_size; + float blur_scale; + int blur_steps; + + bool blur_near_active; + float blur_near_begin; + float blur_near_end; + bool blur_far_active; + + float blur_far_begin; + float blur_far_end; + bool second_pass; + bool half_size; + + bool use_jitter; + float jitter_seed; + uint pad[2]; +} +params; + +//used to work around downsampling filter +#define DEPTH_GAP 0.0 + +const float GOLDEN_ANGLE = 2.39996323; + +//note: uniform pdf rand [0;1[ +float hash12n(vec2 p) { + p = fract(p * vec2(5.3987, 5.4421)); + p += dot(p.yx, p.xy + vec2(21.5351, 14.3137)); + return fract(p.x * p.y * 95.4307); +} diff --git a/servers/rendering/renderer_rd/shaders/bokeh_dof_raster.glsl b/servers/rendering/renderer_rd/shaders/bokeh_dof_raster.glsl new file mode 100644 index 0000000000..a3b3938ee9 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/bokeh_dof_raster.glsl @@ -0,0 +1,253 @@ +/* clang-format off */ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +#include "bokeh_dof_inc.glsl" + +layout(location = 0) out vec2 uv_interp; +/* clang-format on */ + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv_interp = base_arr[gl_VertexIndex]; + + gl_Position = vec4(uv_interp * 2.0 - 1.0, 0.0, 1.0); +} + +/* clang-format off */ +#[fragment] + +#version 450 + +#VERSION_DEFINES + +#include "bokeh_dof_inc.glsl" + +layout(location = 0) in vec2 uv_interp; +/* clang-format on */ + +#ifdef MODE_GEN_BLUR_SIZE +layout(location = 0) out float weight; + +layout(set = 0, binding = 0) uniform sampler2D source_depth; +#else +layout(location = 0) out vec4 frag_color; +#ifdef OUTPUT_WEIGHT +layout(location = 1) out float weight; +#endif + +layout(set = 0, binding = 0) uniform sampler2D source_color; +layout(set = 1, binding = 0) uniform sampler2D source_weight; +#ifdef MODE_COMPOSITE_BOKEH +layout(set = 2, binding = 0) uniform sampler2D original_weight; +#endif +#endif + +//DOF +// Bokeh single pass implementation based on https://tuxedolabs.blogspot.com/2018/05/bokeh-depth-of-field-in-single-pass.html + +#ifdef MODE_GEN_BLUR_SIZE + +float get_depth_at_pos(vec2 uv) { + float depth = textureLod(source_depth, uv, 0.0).x; + if (params.orthogonal) { + depth = ((depth + (params.z_far + params.z_near) / (params.z_far - params.z_near)) * (params.z_far - params.z_near)) / 2.0; + } else { + depth = 2.0 * params.z_near * params.z_far / (params.z_far + params.z_near - depth * (params.z_far - params.z_near)); + } + return depth; +} + +float get_blur_size(float depth) { + if (params.blur_near_active && depth < params.blur_near_begin) { + return -(1.0 - smoothstep(params.blur_near_end, params.blur_near_begin, depth)) * params.blur_size - DEPTH_GAP; //near blur is negative + } + + if (params.blur_far_active && depth > params.blur_far_begin) { + return smoothstep(params.blur_far_begin, params.blur_far_end, depth) * params.blur_size + DEPTH_GAP; + } + + return 0.0; +} + +#endif + +#if defined(MODE_BOKEH_BOX) || defined(MODE_BOKEH_HEXAGONAL) + +vec4 weighted_filter_dir(vec2 dir, vec2 uv, vec2 pixel_size) { + dir *= pixel_size; + vec4 color = texture(source_color, uv); + color.a = texture(source_weight, uv).r; + + vec4 accum = color; + float total = 1.0; + + float blur_scale = params.blur_size / float(params.blur_steps); + + if (params.use_jitter) { + uv += dir * (hash12n(uv + params.jitter_seed) - 0.5); + } + + for (int i = -params.blur_steps; i <= params.blur_steps; i++) { + if (i == 0) { + continue; + } + float radius = float(i) * blur_scale; + vec2 suv = uv + dir * radius; + radius = abs(radius); + + vec4 sample_color = texture(source_color, suv); + sample_color.a = texture(source_weight, suv).r; + float limit; + + if (sample_color.a < color.a) { + limit = abs(sample_color.a); + } else { + limit = abs(color.a); + } + + limit -= DEPTH_GAP; + + float m = smoothstep(radius - 0.5, radius + 0.5, limit); + + accum += mix(color, sample_color, m); + + total += 1.0; + } + + return accum / total; +} + +#endif + +void main() { + vec2 pixel_size = 1.0 / vec2(params.size); + vec2 uv = uv_interp; + +#ifdef MODE_GEN_BLUR_SIZE + uv += pixel_size * 0.5; + float center_depth = get_depth_at_pos(uv); + weight = get_blur_size(center_depth); +#endif + +#ifdef MODE_BOKEH_BOX + //pixel_size*=0.5; //resolution is doubled + if (params.second_pass || !params.half_size) { + uv += pixel_size * 0.5; //half pixel to read centers + } else { + uv += pixel_size * 0.25; //half pixel to read centers from full res + } + + float alpha = texture(source_color, uv).a; // retain this + vec2 dir = (params.second_pass ? vec2(0.0, 1.0) : vec2(1.0, 0.0)); + + vec4 color = weighted_filter_dir(dir, uv, pixel_size); + + frag_color = color; + frag_color.a = alpha; // attempt to retain this in case we have a transparent background, ignored if half_size +#ifdef OUTPUT_WEIGHT + weight = color.a; +#endif + +#endif + +#ifdef MODE_BOKEH_HEXAGONAL + + //pixel_size*=0.5; //resolution is doubled + if (params.second_pass || !params.half_size) { + uv += pixel_size * 0.5; //half pixel to read centers + } else { + uv += pixel_size * 0.25; //half pixel to read centers from full res + } + + float alpha = texture(source_color, uv).a; // retain this + + vec2 dir = (params.second_pass ? normalize(vec2(1.0, 0.577350269189626)) : vec2(0.0, 1.0)); + + vec4 color = weighted_filter_dir(dir, uv, pixel_size); + + if (params.second_pass) { + dir = normalize(vec2(-1.0, 0.577350269189626)); + + vec4 color2 = weighted_filter_dir(dir, uv, pixel_size); + + color.rgb = min(color.rgb, color2.rgb); + color.a = (color.a + color2.a) * 0.5; + } + + frag_color = color; + frag_color.a = alpha; // attempt to retain this in case we have a transparent background, ignored if half_size +#ifdef OUTPUT_WEIGHT + weight = color.a; +#endif + +#endif + +#ifdef MODE_BOKEH_CIRCULAR + if (params.half_size) { + pixel_size *= 0.5; //resolution is doubled + } + + uv += pixel_size * 0.5; //half pixel to read centers + + vec4 color = texture(source_color, uv); + float alpha = color.a; // retain this + color.a = texture(source_weight, uv).r; + + vec4 color_accum = color; + float accum = 1.0; + + float radius = params.blur_scale; + for (float ang = 0.0; radius < params.blur_size; ang += GOLDEN_ANGLE) { + vec2 uv_adj = uv + vec2(cos(ang), sin(ang)) * pixel_size * radius; + + vec4 sample_color = texture(source_color, uv_adj); + sample_color.a = texture(source_weight, uv_adj).r; + + float limit; + + if (sample_color.a < color.a) { + limit = abs(sample_color.a); + } else { + limit = abs(color.a); + } + + limit -= DEPTH_GAP; + + float m = smoothstep(radius - 0.5, radius + 0.5, limit); + color_accum += mix(color_accum / accum, sample_color, m); + accum += 1.0; + + radius += params.blur_scale / radius; + } + + color_accum = color_accum / accum; + + frag_color.rgb = color_accum.rgb; + frag_color.a = alpha; // attempt to retain this in case we have a transparent background, ignored if half_size +#ifdef OUTPUT_WEIGHT + weight = color_accum.a; +#endif + +#endif + +#ifdef MODE_COMPOSITE_BOKEH + frag_color.rgb = texture(source_color, uv).rgb; + + float center_weigth = texture(source_weight, uv).r; + float sample_weight = texture(original_weight, uv).r; + + float mix_amount; + if (sample_weight < center_weigth) { + mix_amount = min(1.0, max(0.0, max(abs(center_weigth), abs(sample_weight)) - DEPTH_GAP)); + } else { + mix_amount = min(1.0, max(0.0, abs(center_weigth) - DEPTH_GAP)); + } + + // let alpha blending take care of mixing + frag_color.a = mix_amount; +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/canvas.glsl b/servers/rendering/renderer_rd/shaders/canvas.glsl index 2186bd174b..2911e8b731 100644 --- a/servers/rendering/renderer_rd/shaders/canvas.glsl +++ b/servers/rendering/renderer_rd/shaders/canvas.glsl @@ -65,7 +65,7 @@ void main() { #elif defined(USE_ATTRIBUTES) vec2 vertex = vertex_attrib; - vec4 color = color_attrib; + vec4 color = color_attrib * draw_data.modulation; vec2 uv = uv_attrib; uvec4 bones = bone_attrib; @@ -458,6 +458,14 @@ void light_blend_compute(uint light_base, vec4 light_color, inout vec3 color) { #endif +float msdf_median(float r, float g, float b, float a) { + return min(max(min(r, g), min(max(r, g), b)), a); +} + +vec2 msdf_map(vec2 value, vec2 in_min, vec2 in_max, vec2 out_min, vec2 out_max) { + return out_min + (out_max - out_min) * (value - in_min) / (in_max - in_min); +} + void main() { vec4 color = color_interp; vec2 uv = uv_interp; @@ -485,7 +493,34 @@ void main() { #endif - color *= texture(sampler2D(color_texture, texture_sampler), uv); +#ifndef USE_PRIMITIVE + if (bool(draw_data.flags & FLAGS_USE_MSDF)) { + float px_range = draw_data.ninepatch_margins.x; + float outline_thickness = draw_data.ninepatch_margins.y; + //float reserved1 = draw_data.ninepatch_margins.z; + //float reserved2 = draw_data.ninepatch_margins.w; + + vec4 msdf_sample = texture(sampler2D(color_texture, texture_sampler), uv); + vec2 msdf_size = vec2(textureSize(sampler2D(color_texture, texture_sampler), 0)); + vec2 dest_size = vec2(1.0) / fwidth(uv); + float px_size = max(0.5 * dot((vec2(px_range) / msdf_size), dest_size), 1.0); + float d = msdf_median(msdf_sample.r, msdf_sample.g, msdf_sample.b, msdf_sample.a) - 0.5; + + if (outline_thickness > 0) { + float cr = clamp(outline_thickness, 0.0, px_range / 2) / px_range; + float a = clamp((d + cr) * px_size, 0.0, 1.0); + color.a = a * color.a; + } else { + float a = clamp(d * px_size + 0.5, 0.0, 1.0); + color.a = a * color.a; + } + + } else { +#else + { +#endif + color *= texture(sampler2D(color_texture, texture_sampler), uv); + } uint light_count = (draw_data.flags >> FLAGS_LIGHT_COUNT_SHIFT) & 0xF; //max 16 lights bool using_light = light_count > 0 || canvas_data.directional_light_count > 0; diff --git a/servers/rendering/renderer_rd/shaders/canvas_uniforms_inc.glsl b/servers/rendering/renderer_rd/shaders/canvas_uniforms_inc.glsl index 451f9b0089..0cff505cae 100644 --- a/servers/rendering/renderer_rd/shaders/canvas_uniforms_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/canvas_uniforms_inc.glsl @@ -24,6 +24,8 @@ #define FLAGS_DEFAULT_NORMAL_MAP_USED (1 << 26) #define FLAGS_DEFAULT_SPECULAR_MAP_USED (1 << 27) +#define FLAGS_USE_MSDF (1 << 28) + #define SAMPLER_NEAREST_CLAMP 0 #define SAMPLER_LINEAR_CLAMP 1 #define SAMPLER_NEAREST_WITH_MIPMAPS_CLAMP 2 diff --git a/servers/rendering/renderer_rd/shaders/cluster_render.glsl b/servers/rendering/renderer_rd/shaders/cluster_render.glsl index da7d189281..6d95722a57 100644 --- a/servers/rendering/renderer_rd/shaders/cluster_render.glsl +++ b/servers/rendering/renderer_rd/shaders/cluster_render.glsl @@ -117,7 +117,7 @@ void main() { uint cluster_thread_group_index; if (!gl_HelperInvocation) { - //http://advances.realtimerendering.com/s2017/2017_Sig_Improved_Culling_final.pdf + //https://advances.realtimerendering.com/s2017/2017_Sig_Improved_Culling_final.pdf uvec4 mask; diff --git a/servers/rendering/renderer_rd/shaders/cube_to_dp.glsl b/servers/rendering/renderer_rd/shaders/cube_to_dp.glsl index dfbce29119..69b895ed29 100644 --- a/servers/rendering/renderer_rd/shaders/cube_to_dp.glsl +++ b/servers/rendering/renderer_rd/shaders/cube_to_dp.glsl @@ -7,8 +7,7 @@ layout(push_constant, binding = 1, std430) uniform Params { float z_far; float z_near; - bool z_flip; - uint pad; + vec2 texel_size; vec4 screen_rect; } params; @@ -35,22 +34,23 @@ layout(set = 0, binding = 0) uniform samplerCube source_cube; layout(push_constant, binding = 1, std430) uniform Params { float z_far; float z_near; - bool z_flip; - uint pad; + vec2 texel_size; vec4 screen_rect; } params; void main() { vec2 uv = uv_interp; + vec2 texel_size = abs(params.texel_size); - vec3 normal = vec3(uv * 2.0 - 1.0, 0.0); + uv = clamp(uv * (1.0 + 2.0 * texel_size) - texel_size, vec2(0.0), vec2(1.0)); - normal.z = 0.5 - 0.5 * ((normal.x * normal.x) + (normal.y * normal.y)); + vec3 normal = vec3(uv * 2.0 - 1.0, 0.0); + normal.z = 0.5 * (1.0 - dot(normal.xy, normal.xy)); // z = 1/2 - 1/2 * (x^2 + y^2) normal = normalize(normal); normal.y = -normal.y; //needs to be flipped to match projection matrix - if (!params.z_flip) { + if (params.texel_size.x >= 0.0) { // Sign is used to encode Z flip normal.z = -normal.z; } diff --git a/servers/rendering/renderer_rd/shaders/cubemap_downsampler.glsl b/servers/rendering/renderer_rd/shaders/cubemap_downsampler.glsl index 9fa84657d1..63f0ce690e 100644 --- a/servers/rendering/renderer_rd/shaders/cubemap_downsampler.glsl +++ b/servers/rendering/renderer_rd/shaders/cubemap_downsampler.glsl @@ -32,53 +32,7 @@ layout(set = 0, binding = 0) uniform samplerCube source_cubemap; layout(rgba16f, set = 1, binding = 0) uniform restrict writeonly imageCube dest_cubemap; -layout(push_constant, binding = 1, std430) uniform Params { - uint face_size; -} -params; - -#define M_PI 3.14159265359 - -void get_dir_0(out vec3 dir, in float u, in float v) { - dir[0] = 1.0; - dir[1] = v; - dir[2] = -u; -} - -void get_dir_1(out vec3 dir, in float u, in float v) { - dir[0] = -1.0; - dir[1] = v; - dir[2] = u; -} - -void get_dir_2(out vec3 dir, in float u, in float v) { - dir[0] = u; - dir[1] = 1.0; - dir[2] = -v; -} - -void get_dir_3(out vec3 dir, in float u, in float v) { - dir[0] = u; - dir[1] = -1.0; - dir[2] = v; -} - -void get_dir_4(out vec3 dir, in float u, in float v) { - dir[0] = u; - dir[1] = v; - dir[2] = 1.0; -} - -void get_dir_5(out vec3 dir, in float u, in float v) { - dir[0] = -u; - dir[1] = v; - dir[2] = -1.0; -} - -float calcWeight(float u, float v) { - float val = u * u + v * v + 1.0; - return val * sqrt(val); -} +#include "cubemap_downsampler_inc.glsl" void main() { uvec3 id = gl_GlobalInvocationID; diff --git a/servers/rendering/renderer_rd/shaders/cubemap_downsampler_inc.glsl b/servers/rendering/renderer_rd/shaders/cubemap_downsampler_inc.glsl new file mode 100644 index 0000000000..b329e67293 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/cubemap_downsampler_inc.glsl @@ -0,0 +1,48 @@ +layout(push_constant, binding = 1, std430) uniform Params { + uint face_size; + uint face_id; // only used in raster shader +} +params; + +#define M_PI 3.14159265359 + +void get_dir_0(out vec3 dir, in float u, in float v) { + dir[0] = 1.0; + dir[1] = v; + dir[2] = -u; +} + +void get_dir_1(out vec3 dir, in float u, in float v) { + dir[0] = -1.0; + dir[1] = v; + dir[2] = u; +} + +void get_dir_2(out vec3 dir, in float u, in float v) { + dir[0] = u; + dir[1] = 1.0; + dir[2] = -v; +} + +void get_dir_3(out vec3 dir, in float u, in float v) { + dir[0] = u; + dir[1] = -1.0; + dir[2] = v; +} + +void get_dir_4(out vec3 dir, in float u, in float v) { + dir[0] = u; + dir[1] = v; + dir[2] = 1.0; +} + +void get_dir_5(out vec3 dir, in float u, in float v) { + dir[0] = -u; + dir[1] = v; + dir[2] = -1.0; +} + +float calcWeight(float u, float v) { + float val = u * u + v * v + 1.0; + return val * sqrt(val); +} diff --git a/servers/rendering/renderer_rd/shaders/cubemap_downsampler_raster.glsl b/servers/rendering/renderer_rd/shaders/cubemap_downsampler_raster.glsl new file mode 100644 index 0000000000..0828ffd921 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/cubemap_downsampler_raster.glsl @@ -0,0 +1,163 @@ +// Copyright 2016 Activision Publishing, Inc. +// +// Permission is hereby granted, free of charge, to any person obtaining +// a copy of this software and associated documentation files (the "Software"), +// to deal in the Software without restriction, including without limitation +// the rights to use, copy, modify, merge, publish, distribute, sublicense, +// and/or sell copies of the Software, and to permit persons to whom the Software +// is furnished to do so, subject to the following conditions: +// +// The above copyright notice and this permission notice shall be included in all +// copies or substantial portions of the Software. +// +// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +// SOFTWARE. + +/* clang-format off */ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +#include "cubemap_downsampler_inc.glsl" + +layout(location = 0) out vec2 uv_interp; +/* clang-format on */ + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv_interp = base_arr[gl_VertexIndex] * float(params.face_size); + gl_Position = vec4(base_arr[gl_VertexIndex] * 2.0 - 1.0, 0.0, 1.0); +} + +/* clang-format off */ +#[fragment] + +#version 450 + +#VERSION_DEFINES + +#include "cubemap_downsampler_inc.glsl" + +layout(set = 0, binding = 0) uniform samplerCube source_cubemap; + +layout(location = 0) in vec2 uv_interp; +layout(location = 0) out vec4 frag_color; +/* clang-format on */ + +void main() { + // Converted from compute shader which uses absolute coordinates. + // Could possibly simplify this + float face_size = float(params.face_size); + + if (uv_interp.x < face_size && uv_interp.y < face_size) { + float inv_face_size = 1.0 / face_size; + + float u0 = (uv_interp.x * 2.0 + 1.0 - 0.75) * inv_face_size - 1.0; + float u1 = (uv_interp.x * 2.0 + 1.0 + 0.75) * inv_face_size - 1.0; + + float v0 = (uv_interp.y * 2.0 + 1.0 - 0.75) * -inv_face_size + 1.0; + float v1 = (uv_interp.y * 2.0 + 1.0 + 0.75) * -inv_face_size + 1.0; + + float weights[4]; + weights[0] = calcWeight(u0, v0); + weights[1] = calcWeight(u1, v0); + weights[2] = calcWeight(u0, v1); + weights[3] = calcWeight(u1, v1); + + const float wsum = 0.5 / (weights[0] + weights[1] + weights[2] + weights[3]); + for (int i = 0; i < 4; i++) { + weights[i] = weights[i] * wsum + .125; + } + + vec3 dir; + vec4 color; + switch (params.face_id) { + case 0: + get_dir_0(dir, u0, v0); + color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0]; + + get_dir_0(dir, u1, v0); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1]; + + get_dir_0(dir, u0, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2]; + + get_dir_0(dir, u1, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3]; + break; + case 1: + get_dir_1(dir, u0, v0); + color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0]; + + get_dir_1(dir, u1, v0); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1]; + + get_dir_1(dir, u0, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2]; + + get_dir_1(dir, u1, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3]; + break; + case 2: + get_dir_2(dir, u0, v0); + color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0]; + + get_dir_2(dir, u1, v0); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1]; + + get_dir_2(dir, u0, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2]; + + get_dir_2(dir, u1, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3]; + break; + case 3: + get_dir_3(dir, u0, v0); + color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0]; + + get_dir_3(dir, u1, v0); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1]; + + get_dir_3(dir, u0, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2]; + + get_dir_3(dir, u1, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3]; + break; + case 4: + get_dir_4(dir, u0, v0); + color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0]; + + get_dir_4(dir, u1, v0); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1]; + + get_dir_4(dir, u0, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2]; + + get_dir_4(dir, u1, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3]; + break; + default: + get_dir_5(dir, u0, v0); + color = textureLod(source_cubemap, normalize(dir), 0.0) * weights[0]; + + get_dir_5(dir, u1, v0); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[1]; + + get_dir_5(dir, u0, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[2]; + + get_dir_5(dir, u1, v1); + color += textureLod(source_cubemap, normalize(dir), 0.0) * weights[3]; + break; + } + frag_color = color; + } +} diff --git a/servers/rendering/renderer_rd/shaders/cubemap_filter_raster.glsl b/servers/rendering/renderer_rd/shaders/cubemap_filter_raster.glsl new file mode 100644 index 0000000000..324d306218 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/cubemap_filter_raster.glsl @@ -0,0 +1,256 @@ +// Copyright 2016 Activision Publishing, Inc. +// +// Permission is hereby granted, free of charge, to any person obtaining +// a copy of this software and associated documentation files (the "Software"), +// to deal in the Software without restriction, including without limitation +// the rights to use, copy, modify, merge, publish, distribute, sublicense, +// and/or sell copies of the Software, and to permit persons to whom the Software +// is furnished to do so, subject to the following conditions: +// +// The above copyright notice and this permission notice shall be included in all +// copies or substantial portions of the Software. +// +// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +// SOFTWARE. + +/* clang-format off */ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +layout(push_constant, binding = 1, std430) uniform Params { + int mip_level; + uint face_id; +} +params; + +layout(location = 0) out vec2 uv_interp; +/* clang-format on */ + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv_interp = base_arr[gl_VertexIndex]; + gl_Position = vec4(base_arr[gl_VertexIndex] * 2.0 - 1.0, 0.0, 1.0); +} + +/* clang-format off */ +#[fragment] + +#version 450 + +#VERSION_DEFINES + +layout(push_constant, binding = 1, std430) uniform Params { + int mip_level; + uint face_id; +} +params; + +layout(set = 0, binding = 0) uniform samplerCube source_cubemap; + +layout(location = 0) in vec2 uv_interp; +layout(location = 0) out vec4 frag_color; + +/* clang-format on */ + +#ifdef USE_HIGH_QUALITY +#define NUM_TAPS 32 +#else +#define NUM_TAPS 8 +#endif + +#define BASE_RESOLUTION 128 + +#ifdef USE_HIGH_QUALITY +layout(set = 1, binding = 0, std430) buffer restrict readonly Data { + vec4[7][5][3][24] coeffs; +} +data; +#else +layout(set = 1, binding = 0, std430) buffer restrict readonly Data { + vec4[7][5][6] coeffs; +} +data; +#endif + +void get_dir(out vec3 dir, in vec2 uv, in uint face) { + switch (face) { + case 0: + dir = vec3(1.0, uv[1], -uv[0]); + break; + case 1: + dir = vec3(-1.0, uv[1], uv[0]); + break; + case 2: + dir = vec3(uv[0], 1.0, -uv[1]); + break; + case 3: + dir = vec3(uv[0], -1.0, uv[1]); + break; + case 4: + dir = vec3(uv[0], uv[1], 1.0); + break; + default: + dir = vec3(-uv[0], uv[1], -1.0); + break; + } +} + +void main() { + // determine dir / pos for the texel + vec3 dir, adir, frameZ; + { + vec2 uv; + uv.x = uv_interp.x; + uv.y = 1.0 - uv_interp.y; + uv = uv * 2.0 - 1.0; + + get_dir(dir, uv, params.face_id); + frameZ = normalize(dir); + + adir = abs(dir); + } + + // determine which texel this is + // NOTE (macOS/MoltenVK): Do not rename, "level" variable name conflicts with the Metal "level(float lod)" mipmap sampling function name. + int mip_level = 0; + + if (params.mip_level < 0) { + // return as is + frag_color.rgb = textureLod(source_cubemap, frameZ, 0.0).rgb; + frag_color.a = 1.0; + return; + } else if (params.mip_level > 6) { + // maximum level + mip_level = 6; + } else { + mip_level = params.mip_level; + } + + // GGX gather colors + vec4 color = vec4(0.0); + for (int axis = 0; axis < 3; axis++) { + const int otherAxis0 = 1 - (axis & 1) - (axis >> 1); + const int otherAxis1 = 2 - (axis >> 1); + + float frameweight = (max(adir[otherAxis0], adir[otherAxis1]) - .75) / .25; + if (frameweight > 0.0) { + // determine frame + vec3 UpVector; + switch (axis) { + case 0: + UpVector = vec3(1, 0, 0); + break; + case 1: + UpVector = vec3(0, 1, 0); + break; + default: + UpVector = vec3(0, 0, 1); + break; + } + + vec3 frameX = normalize(cross(UpVector, frameZ)); + vec3 frameY = cross(frameZ, frameX); + + // calculate parametrization for polynomial + float Nx = dir[otherAxis0]; + float Ny = dir[otherAxis1]; + float Nz = adir[axis]; + + float NmaxXY = max(abs(Ny), abs(Nx)); + Nx /= NmaxXY; + Ny /= NmaxXY; + + float theta; + if (Ny < Nx) { + if (Ny <= -0.999) + theta = Nx; + else + theta = Ny; + } else { + if (Ny >= 0.999) + theta = -Nx; + else + theta = -Ny; + } + + float phi; + if (Nz <= -0.999) + phi = -NmaxXY; + else if (Nz >= 0.999) + phi = NmaxXY; + else + phi = Nz; + + float theta2 = theta * theta; + float phi2 = phi * phi; + + // sample + for (int iSuperTap = 0; iSuperTap < NUM_TAPS / 4; iSuperTap++) { + const int index = (NUM_TAPS / 4) * axis + iSuperTap; + +#ifdef USE_HIGH_QUALITY + vec4 coeffsDir0[3]; + vec4 coeffsDir1[3]; + vec4 coeffsDir2[3]; + vec4 coeffsLevel[3]; + vec4 coeffsWeight[3]; + + for (int iCoeff = 0; iCoeff < 3; iCoeff++) { + coeffsDir0[iCoeff] = data.coeffs[mip_level][0][iCoeff][index]; + coeffsDir1[iCoeff] = data.coeffs[mip_level][1][iCoeff][index]; + coeffsDir2[iCoeff] = data.coeffs[mip_level][2][iCoeff][index]; + coeffsLevel[iCoeff] = data.coeffs[mip_level][3][iCoeff][index]; + coeffsWeight[iCoeff] = data.coeffs[mip_level][4][iCoeff][index]; + } + + for (int iSubTap = 0; iSubTap < 4; iSubTap++) { + // determine sample attributes (dir, weight, mip_level) + vec3 sample_dir = frameX * (coeffsDir0[0][iSubTap] + coeffsDir0[1][iSubTap] * theta2 + coeffsDir0[2][iSubTap] * phi2) + frameY * (coeffsDir1[0][iSubTap] + coeffsDir1[1][iSubTap] * theta2 + coeffsDir1[2][iSubTap] * phi2) + frameZ * (coeffsDir2[0][iSubTap] + coeffsDir2[1][iSubTap] * theta2 + coeffsDir2[2][iSubTap] * phi2); + + float sample_level = coeffsLevel[0][iSubTap] + coeffsLevel[1][iSubTap] * theta2 + coeffsLevel[2][iSubTap] * phi2; + + float sample_weight = coeffsWeight[0][iSubTap] + coeffsWeight[1][iSubTap] * theta2 + coeffsWeight[2][iSubTap] * phi2; +#else + vec4 coeffsDir0 = data.coeffs[mip_level][0][index]; + vec4 coeffsDir1 = data.coeffs[mip_level][1][index]; + vec4 coeffsDir2 = data.coeffs[mip_level][2][index]; + vec4 coeffsLevel = data.coeffs[mip_level][3][index]; + vec4 coeffsWeight = data.coeffs[mip_level][4][index]; + + for (int iSubTap = 0; iSubTap < 4; iSubTap++) { + // determine sample attributes (dir, weight, mip_level) + vec3 sample_dir = frameX * coeffsDir0[iSubTap] + frameY * coeffsDir1[iSubTap] + frameZ * coeffsDir2[iSubTap]; + + float sample_level = coeffsLevel[iSubTap]; + + float sample_weight = coeffsWeight[iSubTap]; +#endif + + sample_weight *= frameweight; + + // adjust for jacobian + sample_dir /= max(abs(sample_dir[0]), max(abs(sample_dir[1]), abs(sample_dir[2]))); + sample_level += 0.75 * log2(dot(sample_dir, sample_dir)); + // sample cubemap + color.xyz += textureLod(source_cubemap, normalize(sample_dir), sample_level).xyz * sample_weight; + color.w += sample_weight; + } + } + } + } + color /= color.w; + + // write color + color.xyz = max(vec3(0.0), color.xyz); + color.w = 1.0; + + frag_color = color; +} diff --git a/servers/rendering/renderer_rd/shaders/cubemap_roughness.glsl b/servers/rendering/renderer_rd/shaders/cubemap_roughness.glsl index ce7c03c1d4..28f4dc59ec 100644 --- a/servers/rendering/renderer_rd/shaders/cubemap_roughness.glsl +++ b/servers/rendering/renderer_rd/shaders/cubemap_roughness.glsl @@ -12,100 +12,7 @@ layout(set = 0, binding = 0) uniform samplerCube source_cube; layout(rgba16f, set = 1, binding = 0) uniform restrict writeonly imageCube dest_cubemap; -layout(push_constant, binding = 1, std430) uniform Params { - uint face_id; - uint sample_count; - float roughness; - bool use_direct_write; - float face_size; -} -params; - -#define M_PI 3.14159265359 - -vec3 texelCoordToVec(vec2 uv, uint faceID) { - mat3 faceUvVectors[6]; - - // -x - faceUvVectors[1][0] = vec3(0.0, 0.0, 1.0); // u -> +z - faceUvVectors[1][1] = vec3(0.0, -1.0, 0.0); // v -> -y - faceUvVectors[1][2] = vec3(-1.0, 0.0, 0.0); // -x face - - // +x - faceUvVectors[0][0] = vec3(0.0, 0.0, -1.0); // u -> -z - faceUvVectors[0][1] = vec3(0.0, -1.0, 0.0); // v -> -y - faceUvVectors[0][2] = vec3(1.0, 0.0, 0.0); // +x face - - // -y - faceUvVectors[3][0] = vec3(1.0, 0.0, 0.0); // u -> +x - faceUvVectors[3][1] = vec3(0.0, 0.0, -1.0); // v -> -z - faceUvVectors[3][2] = vec3(0.0, -1.0, 0.0); // -y face - - // +y - faceUvVectors[2][0] = vec3(1.0, 0.0, 0.0); // u -> +x - faceUvVectors[2][1] = vec3(0.0, 0.0, 1.0); // v -> +z - faceUvVectors[2][2] = vec3(0.0, 1.0, 0.0); // +y face - - // -z - faceUvVectors[5][0] = vec3(-1.0, 0.0, 0.0); // u -> -x - faceUvVectors[5][1] = vec3(0.0, -1.0, 0.0); // v -> -y - faceUvVectors[5][2] = vec3(0.0, 0.0, -1.0); // -z face - - // +z - faceUvVectors[4][0] = vec3(1.0, 0.0, 0.0); // u -> +x - faceUvVectors[4][1] = vec3(0.0, -1.0, 0.0); // v -> -y - faceUvVectors[4][2] = vec3(0.0, 0.0, 1.0); // +z face - - // out = u * s_faceUv[0] + v * s_faceUv[1] + s_faceUv[2]. - vec3 result = (faceUvVectors[faceID][0] * uv.x) + (faceUvVectors[faceID][1] * uv.y) + faceUvVectors[faceID][2]; - return normalize(result); -} - -vec3 ImportanceSampleGGX(vec2 Xi, float Roughness, vec3 N) { - float a = Roughness * Roughness; // DISNEY'S ROUGHNESS [see Burley'12 siggraph] - - // Compute distribution direction - float Phi = 2.0 * M_PI * Xi.x; - float CosTheta = sqrt((1.0 - Xi.y) / (1.0 + (a * a - 1.0) * Xi.y)); - float SinTheta = sqrt(1.0 - CosTheta * CosTheta); - - // Convert to spherical direction - vec3 H; - H.x = SinTheta * cos(Phi); - H.y = SinTheta * sin(Phi); - H.z = CosTheta; - - vec3 UpVector = abs(N.z) < 0.999 ? vec3(0.0, 0.0, 1.0) : vec3(1.0, 0.0, 0.0); - vec3 TangentX = normalize(cross(UpVector, N)); - vec3 TangentY = cross(N, TangentX); - - // Tangent to world space - return TangentX * H.x + TangentY * H.y + N * H.z; -} - -// http://graphicrants.blogspot.com.au/2013/08/specular-brdf-reference.html -float GGX(float NdotV, float a) { - float k = a / 2.0; - return NdotV / (NdotV * (1.0 - k) + k); -} - -// http://graphicrants.blogspot.com.au/2013/08/specular-brdf-reference.html -float G_Smith(float a, float nDotV, float nDotL) { - return GGX(nDotL, a * a) * GGX(nDotV, a * a); -} - -float radicalInverse_VdC(uint bits) { - bits = (bits << 16u) | (bits >> 16u); - bits = ((bits & 0x55555555u) << 1u) | ((bits & 0xAAAAAAAAu) >> 1u); - bits = ((bits & 0x33333333u) << 2u) | ((bits & 0xCCCCCCCCu) >> 2u); - bits = ((bits & 0x0F0F0F0Fu) << 4u) | ((bits & 0xF0F0F0F0u) >> 4u); - bits = ((bits & 0x00FF00FFu) << 8u) | ((bits & 0xFF00FF00u) >> 8u); - return float(bits) * 2.3283064365386963e-10; // / 0x100000000 -} - -vec2 Hammersley(uint i, uint N) { - return vec2(float(i) / float(N), radicalInverse_VdC(i)); -} +#include "cubemap_roughness_inc.glsl" void main() { uvec3 id = gl_GlobalInvocationID; diff --git a/servers/rendering/renderer_rd/shaders/cubemap_roughness_inc.glsl b/servers/rendering/renderer_rd/shaders/cubemap_roughness_inc.glsl new file mode 100644 index 0000000000..be12be5dec --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/cubemap_roughness_inc.glsl @@ -0,0 +1,94 @@ +#define M_PI 3.14159265359 + +layout(push_constant, binding = 1, std430) uniform Params { + uint face_id; + uint sample_count; + float roughness; + bool use_direct_write; + float face_size; +} +params; + +vec3 texelCoordToVec(vec2 uv, uint faceID) { + mat3 faceUvVectors[6]; + + // -x + faceUvVectors[1][0] = vec3(0.0, 0.0, 1.0); // u -> +z + faceUvVectors[1][1] = vec3(0.0, -1.0, 0.0); // v -> -y + faceUvVectors[1][2] = vec3(-1.0, 0.0, 0.0); // -x face + + // +x + faceUvVectors[0][0] = vec3(0.0, 0.0, -1.0); // u -> -z + faceUvVectors[0][1] = vec3(0.0, -1.0, 0.0); // v -> -y + faceUvVectors[0][2] = vec3(1.0, 0.0, 0.0); // +x face + + // -y + faceUvVectors[3][0] = vec3(1.0, 0.0, 0.0); // u -> +x + faceUvVectors[3][1] = vec3(0.0, 0.0, -1.0); // v -> -z + faceUvVectors[3][2] = vec3(0.0, -1.0, 0.0); // -y face + + // +y + faceUvVectors[2][0] = vec3(1.0, 0.0, 0.0); // u -> +x + faceUvVectors[2][1] = vec3(0.0, 0.0, 1.0); // v -> +z + faceUvVectors[2][2] = vec3(0.0, 1.0, 0.0); // +y face + + // -z + faceUvVectors[5][0] = vec3(-1.0, 0.0, 0.0); // u -> -x + faceUvVectors[5][1] = vec3(0.0, -1.0, 0.0); // v -> -y + faceUvVectors[5][2] = vec3(0.0, 0.0, -1.0); // -z face + + // +z + faceUvVectors[4][0] = vec3(1.0, 0.0, 0.0); // u -> +x + faceUvVectors[4][1] = vec3(0.0, -1.0, 0.0); // v -> -y + faceUvVectors[4][2] = vec3(0.0, 0.0, 1.0); // +z face + + // out = u * s_faceUv[0] + v * s_faceUv[1] + s_faceUv[2]. + vec3 result = (faceUvVectors[faceID][0] * uv.x) + (faceUvVectors[faceID][1] * uv.y) + faceUvVectors[faceID][2]; + return normalize(result); +} + +vec3 ImportanceSampleGGX(vec2 Xi, float Roughness, vec3 N) { + float a = Roughness * Roughness; // DISNEY'S ROUGHNESS [see Burley'12 siggraph] + + // Compute distribution direction + float Phi = 2.0 * M_PI * Xi.x; + float CosTheta = sqrt((1.0 - Xi.y) / (1.0 + (a * a - 1.0) * Xi.y)); + float SinTheta = sqrt(1.0 - CosTheta * CosTheta); + + // Convert to spherical direction + vec3 H; + H.x = SinTheta * cos(Phi); + H.y = SinTheta * sin(Phi); + H.z = CosTheta; + + vec3 UpVector = abs(N.z) < 0.999 ? vec3(0.0, 0.0, 1.0) : vec3(1.0, 0.0, 0.0); + vec3 TangentX = normalize(cross(UpVector, N)); + vec3 TangentY = cross(N, TangentX); + + // Tangent to world space + return TangentX * H.x + TangentY * H.y + N * H.z; +} + +// https://graphicrants.blogspot.com.au/2013/08/specular-brdf-reference.html +float GGX(float NdotV, float a) { + float k = a / 2.0; + return NdotV / (NdotV * (1.0 - k) + k); +} + +// https://graphicrants.blogspot.com.au/2013/08/specular-brdf-reference.html +float G_Smith(float a, float nDotV, float nDotL) { + return GGX(nDotL, a * a) * GGX(nDotV, a * a); +} + +float radicalInverse_VdC(uint bits) { + bits = (bits << 16u) | (bits >> 16u); + bits = ((bits & 0x55555555u) << 1u) | ((bits & 0xAAAAAAAAu) >> 1u); + bits = ((bits & 0x33333333u) << 2u) | ((bits & 0xCCCCCCCCu) >> 2u); + bits = ((bits & 0x0F0F0F0Fu) << 4u) | ((bits & 0xF0F0F0F0u) >> 4u); + bits = ((bits & 0x00FF00FFu) << 8u) | ((bits & 0xFF00FF00u) >> 8u); + return float(bits) * 2.3283064365386963e-10; // / 0x100000000 +} + +vec2 Hammersley(uint i, uint N) { + return vec2(float(i) / float(N), radicalInverse_VdC(i)); +} diff --git a/servers/rendering/renderer_rd/shaders/cubemap_roughness_raster.glsl b/servers/rendering/renderer_rd/shaders/cubemap_roughness_raster.glsl new file mode 100644 index 0000000000..2570308816 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/cubemap_roughness_raster.glsl @@ -0,0 +1,63 @@ +/* clang-format off */ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +#include "cubemap_roughness_inc.glsl" + +layout(location = 0) out vec2 uv_interp; +/* clang-format on */ + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv_interp = base_arr[gl_VertexIndex]; + gl_Position = vec4(uv_interp * 2.0 - 1.0, 0.0, 1.0); +} + +/* clang-format off */ +#[fragment] + +#version 450 + +#VERSION_DEFINES + +#include "cubemap_roughness_inc.glsl" + +layout(location = 0) in vec2 uv_interp; + +layout(set = 0, binding = 0) uniform samplerCube source_cube; + +layout(location = 0) out vec4 frag_color; +/* clang-format on */ + +void main() { + vec3 N = texelCoordToVec(uv_interp * 2.0 - 1.0, params.face_id); + + //vec4 color = color_interp; + + if (params.use_direct_write) { + frag_color = vec4(texture(source_cube, N).rgb, 1.0); + } else { + vec4 sum = vec4(0.0, 0.0, 0.0, 0.0); + + for (uint sampleNum = 0u; sampleNum < params.sample_count; sampleNum++) { + vec2 xi = Hammersley(sampleNum, params.sample_count); + + vec3 H = ImportanceSampleGGX(xi, params.roughness, N); + vec3 V = N; + vec3 L = (2.0 * dot(V, H) * H - V); + + float ndotl = clamp(dot(N, L), 0.0, 1.0); + + if (ndotl > 0.0) { + sum.rgb += textureLod(source_cube, L, 0.0).rgb * ndotl; + sum.a += ndotl; + } + } + sum /= sum.a; + + frag_color = vec4(sum.rgb, 1.0); + } +} diff --git a/servers/rendering/renderer_rd/shaders/decal_data_inc.glsl b/servers/rendering/renderer_rd/shaders/decal_data_inc.glsl index ccaad13311..158096d3c7 100644 --- a/servers/rendering/renderer_rd/shaders/decal_data_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/decal_data_inc.glsl @@ -1,18 +1,18 @@ struct DecalData { - mat4 xform; //to decal transform - vec3 inv_extents; - float albedo_mix; - vec4 albedo_rect; - vec4 normal_rect; - vec4 orm_rect; - vec4 emission_rect; - vec4 modulate; - float emission_energy; + highp mat4 xform; //to decal transform + highp vec3 inv_extents; + mediump float albedo_mix; + highp vec4 albedo_rect; + highp vec4 normal_rect; + highp vec4 orm_rect; + highp vec4 emission_rect; + highp vec4 modulate; + mediump float emission_energy; uint mask; - float upper_fade; - float lower_fade; - mat3x4 normal_xform; - vec3 normal; - float normal_fade; + mediump float upper_fade; + mediump float lower_fade; + mediump mat3x4 normal_xform; + mediump vec3 normal; + mediump float normal_fade; }; diff --git a/servers/rendering/renderer_rd/shaders/gi.glsl b/servers/rendering/renderer_rd/shaders/gi.glsl index 3977f4efa0..60c881881d 100644 --- a/servers/rendering/renderer_rd/shaders/gi.glsl +++ b/servers/rendering/renderer_rd/shaders/gi.glsl @@ -77,9 +77,9 @@ struct VoxelGIData { bool blend_ambient; uint texture_slot; - float anisotropy_strength; - float ambient_occlusion; - float ambient_occlusion_size; + uint pad0; + uint pad1; + uint pad2; uint mipmaps; }; @@ -551,27 +551,6 @@ void voxel_gi_compute(uint index, vec3 position, vec3 normal, vec3 ref_vec, mat3 } } - if (voxel_gi_instances.data[index].ambient_occlusion > 0.001) { - float size = 1.0 + voxel_gi_instances.data[index].ambient_occlusion_size * 7.0; - - float taps, blend; - blend = modf(size, taps); - float ao = 0.0; - for (float i = 1.0; i <= taps; i++) { - vec3 ofs = (position + normal * (i * 0.5 + 1.0)) * cell_size; - ao += textureLod(sampler3D(voxel_gi_textures[index], linear_sampler_with_mipmaps), ofs, i - 1.0).a * i; - } - - if (blend > 0.001) { - vec3 ofs = (position + normal * ((taps + 1.0) * 0.5 + 1.0)) * cell_size; - ao += textureLod(sampler3D(voxel_gi_textures[index], linear_sampler_with_mipmaps), ofs, taps).a * (taps + 1.0) * blend; - } - - ao = 1.0 - min(1.0, ao); - - light.rgb = mix(params.ao_color, light.rgb, mix(1.0, ao, voxel_gi_instances.data[index].ambient_occlusion)); - } - light.rgb *= voxel_gi_instances.data[index].dynamic_range; if (!voxel_gi_instances.data[index].blend_ambient) { light.a = 1.0; diff --git a/servers/rendering/renderer_rd/shaders/giprobe_write.glsl b/servers/rendering/renderer_rd/shaders/giprobe_write.glsl index 5dc2d08a3b..25d87ca45d 100644 --- a/servers/rendering/renderer_rd/shaders/giprobe_write.glsl +++ b/servers/rendering/renderer_rd/shaders/giprobe_write.glsl @@ -202,12 +202,7 @@ void main() { vec3 emission = vec3(ivec3(cell_data.data[cell_index].emission & 0x3FF, (cell_data.data[cell_index].emission >> 10) & 0x7FF, cell_data.data[cell_index].emission >> 21)) * params.emission_scale; vec4 normal = unpackSnorm4x8(cell_data.data[cell_index].normal); -#ifdef MODE_ANISOTROPIC - vec3 accum[6] = vec3[](vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0)); - const vec3 accum_dirs[6] = vec3[](vec3(1.0, 0.0, 0.0), vec3(-1.0, 0.0, 0.0), vec3(0.0, 1.0, 0.0), vec3(0.0, -1.0, 0.0), vec3(0.0, 0.0, 1.0), vec3(0.0, 0.0, -1.0)); -#else vec3 accum = vec3(0.0); -#endif for (uint i = 0; i < params.light_count; i++) { float attenuation; @@ -242,77 +237,35 @@ void main() { vec3 light = lights.data[i].color * albedo.rgb * attenuation * lights.data[i].energy; -#ifdef MODE_ANISOTROPIC - for (uint j = 0; j < 6; j++) { - accum[j] += max(0.0, dot(accum_dir, -light_dir)) * light + emission; - } -#else if (length(normal.xyz) > 0.2) { accum += max(0.0, dot(normal.xyz, -light_dir)) * light + emission; } else { //all directions accum += light + emission; } -#endif } -#ifdef MODE_ANISOTROPIC - - output.data[cell_index * 6 + 0] = vec4(accum[0], 0.0); - output.data[cell_index * 6 + 1] = vec4(accum[1], 0.0); - output.data[cell_index * 6 + 2] = vec4(accum[2], 0.0); - output.data[cell_index * 6 + 3] = vec4(accum[3], 0.0); - output.data[cell_index * 6 + 4] = vec4(accum[4], 0.0); - output.data[cell_index * 6 + 5] = vec4(accum[5], 0.0); -#else output.data[cell_index] = vec4(accum, 0.0); -#endif - #endif //MODE_COMPUTE_LIGHT #ifdef MODE_UPDATE_MIPMAPS { -#ifdef MODE_ANISOTROPIC - vec3 light_accum[6] = vec3[](vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0)); -#else vec3 light_accum = vec3(0.0); -#endif float count = 0.0; for (uint i = 0; i < 8; i++) { uint child_index = cell_children.data[cell_index].children[i]; if (child_index == NO_CHILDREN) { continue; } -#ifdef MODE_ANISOTROPIC - light_accum[1] += output.data[child_index * 6 + 0].rgb; - light_accum[2] += output.data[child_index * 6 + 1].rgb; - light_accum[3] += output.data[child_index * 6 + 2].rgb; - light_accum[4] += output.data[child_index * 6 + 3].rgb; - light_accum[5] += output.data[child_index * 6 + 4].rgb; - light_accum[6] += output.data[child_index * 6 + 5].rgb; - -#else light_accum += output.data[child_index].rgb; -#endif - count += 1.0; } float divisor = mix(8.0, count, params.propagation); -#ifdef MODE_ANISOTROPIC - output.data[cell_index * 6 + 0] = vec4(light_accum[0] / divisor, 0.0); - output.data[cell_index * 6 + 1] = vec4(light_accum[1] / divisor, 0.0); - output.data[cell_index * 6 + 2] = vec4(light_accum[2] / divisor, 0.0); - output.data[cell_index * 6 + 3] = vec4(light_accum[3] / divisor, 0.0); - output.data[cell_index * 6 + 4] = vec4(light_accum[4] / divisor, 0.0); - output.data[cell_index * 6 + 5] = vec4(light_accum[5] / divisor, 0.0); - -#else output.data[cell_index] = vec4(light_accum / divisor, 0.0); -#endif } #endif diff --git a/servers/rendering/renderer_rd/shaders/light_data_inc.glsl b/servers/rendering/renderer_rd/shaders/light_data_inc.glsl index 2fce258cff..fdc7729338 100644 --- a/servers/rendering/renderer_rd/shaders/light_data_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/light_data_inc.glsl @@ -3,31 +3,31 @@ #define LIGHT_BAKE_STATIC 2 struct LightData { //this structure needs to be as packed as possible - vec3 position; - float inv_radius; + highp vec3 position; + highp float inv_radius; - vec3 direction; - float size; + mediump vec3 direction; + highp float size; - vec3 color; - float attenuation; + mediump vec3 color; + mediump float attenuation; - float cone_attenuation; - float cone_angle; - float specular_amount; + mediump float cone_attenuation; + mediump float cone_angle; + mediump float specular_amount; bool shadow_enabled; - vec4 atlas_rect; // rect in the shadow atlas - mat4 shadow_matrix; - float shadow_bias; - float shadow_normal_bias; - float transmittance_bias; - float soft_shadow_size; // for spot, it's the size in uv coordinates of the light, for omni it's the span angle - float soft_shadow_scale; // scales the shadow kernel for blurrier shadows + highp vec4 atlas_rect; // rect in the shadow atlas + highp mat4 shadow_matrix; + highp float shadow_bias; + highp float shadow_normal_bias; + highp float transmittance_bias; + highp float soft_shadow_size; // for spot, it's the size in uv coordinates of the light, for omni it's the span angle + highp float soft_shadow_scale; // scales the shadow kernel for blurrier shadows uint mask; - float shadow_volumetric_fog_fade; + mediump float shadow_volumetric_fog_fade; uint bake_mode; - vec4 projector_rect; //projector rect in srgb decal atlas + highp vec4 projector_rect; //projector rect in srgb decal atlas }; #define REFLECTION_AMBIENT_DISABLED 0 @@ -35,53 +35,53 @@ struct LightData { //this structure needs to be as packed as possible #define REFLECTION_AMBIENT_COLOR 2 struct ReflectionData { - vec3 box_extents; - float index; - vec3 box_offset; + highp vec3 box_extents; + mediump float index; + highp vec3 box_offset; uint mask; - vec3 ambient; // ambient color - float intensity; + mediump vec3 ambient; // ambient color + mediump float intensity; bool exterior; bool box_project; uint ambient_mode; uint pad; //0-8 is intensity,8-9 is ambient, mode - mat4 local_matrix; // up to here for spot and omni, rest is for directional + highp mat4 local_matrix; // up to here for spot and omni, rest is for directional // notes: for ambientblend, use distance to edge to blend between already existing global environment }; struct DirectionalLightData { - vec3 direction; - float energy; - vec3 color; - float size; - float specular; + mediump vec3 direction; + mediump float energy; + mediump vec3 color; + mediump float size; + mediump float specular; uint mask; - float softshadow_angle; - float soft_shadow_scale; + highp float softshadow_angle; + highp float soft_shadow_scale; bool blend_splits; bool shadow_enabled; - float fade_from; - float fade_to; + highp float fade_from; + highp float fade_to; uvec2 pad; uint bake_mode; - float shadow_volumetric_fog_fade; - vec4 shadow_bias; - vec4 shadow_normal_bias; - vec4 shadow_transmittance_bias; - vec4 shadow_z_range; - vec4 shadow_range_begin; - vec4 shadow_split_offsets; - mat4 shadow_matrix1; - mat4 shadow_matrix2; - mat4 shadow_matrix3; - mat4 shadow_matrix4; - vec4 shadow_color1; - vec4 shadow_color2; - vec4 shadow_color3; - vec4 shadow_color4; - vec2 uv_scale1; - vec2 uv_scale2; - vec2 uv_scale3; - vec2 uv_scale4; + mediump float shadow_volumetric_fog_fade; + highp vec4 shadow_bias; + highp vec4 shadow_normal_bias; + highp vec4 shadow_transmittance_bias; + highp vec4 shadow_z_range; + highp vec4 shadow_range_begin; + highp vec4 shadow_split_offsets; + highp mat4 shadow_matrix1; + highp mat4 shadow_matrix2; + highp mat4 shadow_matrix3; + highp mat4 shadow_matrix4; + mediump vec4 shadow_color1; + mediump vec4 shadow_color2; + mediump vec4 shadow_color3; + mediump vec4 shadow_color4; + highp vec2 uv_scale1; + highp vec2 uv_scale2; + highp vec2 uv_scale3; + highp vec2 uv_scale4; }; diff --git a/servers/rendering/renderer_rd/shaders/luminance_reduce_raster.glsl b/servers/rendering/renderer_rd/shaders/luminance_reduce_raster.glsl new file mode 100644 index 0000000000..29ebd74a90 --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/luminance_reduce_raster.glsl @@ -0,0 +1,74 @@ +/* clang-format off */ +#[vertex] + +#version 450 + +#VERSION_DEFINES + +#include "luminance_reduce_raster_inc.glsl" + +layout(location = 0) out vec2 uv_interp; +/* clang-format on */ + +void main() { + vec2 base_arr[4] = vec2[](vec2(0.0, 0.0), vec2(0.0, 1.0), vec2(1.0, 1.0), vec2(1.0, 0.0)); + uv_interp = base_arr[gl_VertexIndex]; + + gl_Position = vec4(uv_interp * 2.0 - 1.0, 0.0, 1.0); +} + +/* clang-format off */ +#[fragment] + +#version 450 + +#VERSION_DEFINES + +#include "luminance_reduce_raster_inc.glsl" + +layout(location = 0) in vec2 uv_interp; +/* clang-format on */ + +layout(set = 0, binding = 0) uniform sampler2D source_exposure; + +#ifdef FINAL_PASS +layout(set = 1, binding = 0) uniform sampler2D prev_luminance; +#endif + +layout(location = 0) out highp float luminance; + +void main() { + ivec2 dest_pos = ivec2(uv_interp * settings.dest_size); + ivec2 src_pos = ivec2(uv_interp * settings.source_size); + + ivec2 next_pos = (dest_pos + ivec2(1)) * settings.source_size / settings.dest_size; + next_pos = max(next_pos, src_pos + ivec2(1)); //so it at least reads one pixel + + highp vec3 source_color = vec3(0.0); + for (int i = src_pos.x; i < next_pos.x; i++) { + for (int j = src_pos.y; j < next_pos.y; j++) { + source_color += texelFetch(source_exposure, ivec2(i, j), 0).rgb; + } + } + + source_color /= float((next_pos.x - src_pos.x) * (next_pos.y - src_pos.y)); + +#ifdef FIRST_PASS + luminance = max(source_color.r, max(source_color.g, source_color.b)); + + // This formula should be more "accurate" but gave an overexposed result when testing. + // Leaving it here so we can revisit it if we want. + // luminance = source_color.r * 0.21 + source_color.g * 0.71 + source_color.b * 0.07; +#else + luminance = source_color.r; +#endif + +#ifdef FINAL_PASS + // Obtain our target luminance + luminance = clamp(luminance, settings.min_luminance, settings.max_luminance); + + // Now smooth to our transition + highp float prev_lum = texelFetch(prev_luminance, ivec2(0, 0), 0).r; //1 pixel previous luminance + luminance = prev_lum + (luminance - prev_lum) * clamp(settings.exposure_adjust, 0.0, 1.0); +#endif +} diff --git a/servers/rendering/renderer_rd/shaders/luminance_reduce_raster_inc.glsl b/servers/rendering/renderer_rd/shaders/luminance_reduce_raster_inc.glsl new file mode 100644 index 0000000000..3cde9923fa --- /dev/null +++ b/servers/rendering/renderer_rd/shaders/luminance_reduce_raster_inc.glsl @@ -0,0 +1,11 @@ + +layout(push_constant, binding = 1, std430) uniform PushConstant { + ivec2 source_size; + ivec2 dest_size; + + float exposure_adjust; + float min_luminance; + float max_luminance; + uint pad1; +} +settings; diff --git a/servers/rendering/renderer_rd/shaders/particles_copy.glsl b/servers/rendering/renderer_rd/shaders/particles_copy.glsl index 4dceeea995..e88e68b511 100644 --- a/servers/rendering/renderer_rd/shaders/particles_copy.glsl +++ b/servers/rendering/renderer_rd/shaders/particles_copy.glsl @@ -138,7 +138,7 @@ void main() { if (bool(particles.data[particle].flags & PARTICLE_FLAG_ACTIVE) || bool(particles.data[particle].flags & PARTICLE_FLAG_TRAILED)) { txform = particles.data[particle].xform; if (params.trail_size > 1) { - // since the steps dont fit precisely in the history frames, must do a tiny bit of + // Since the steps don't fit precisely in the history frames, must do a tiny bit of // interpolation to get them close to their intended location. uint part_ofs = particle % params.trail_size; float natural_ofs = fract((float(part_ofs) / float(params.trail_size)) * float(params.trail_total)) * params.frame_delta; diff --git a/servers/rendering/renderer_rd/shaders/resolve.glsl b/servers/rendering/renderer_rd/shaders/resolve.glsl index a4610e081c..fecf812a8c 100644 --- a/servers/rendering/renderer_rd/shaders/resolve.glsl +++ b/servers/rendering/renderer_rd/shaders/resolve.glsl @@ -6,6 +6,11 @@ layout(local_size_x = 8, local_size_y = 8, local_size_z = 1) in; +#ifdef MODE_RESOLVE_DEPTH +layout(set = 0, binding = 0) uniform sampler2DMS source_depth; +layout(r32f, set = 1, binding = 0) uniform restrict writeonly image2D dest_depth; +#endif + #ifdef MODE_RESOLVE_GI layout(set = 0, binding = 0) uniform sampler2DMS source_depth; layout(set = 0, binding = 1) uniform sampler2DMS source_normal_roughness; @@ -34,6 +39,17 @@ void main() { return; } +#ifdef MODE_RESOLVE_DEPTH + + float depth_avg = 0.0; + for (int i = 0; i < params.sample_count; i++) { + depth_avg += texelFetch(source_depth, pos, i).r; + } + depth_avg /= float(params.sample_count); + imageStore(dest_depth, pos, vec4(depth_avg)); + +#endif + #ifdef MODE_RESOLVE_GI float best_depth = 1e20; diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl index ce02d5c4d0..edbe1031b7 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_clustered.glsl @@ -118,7 +118,7 @@ void main() { mat3 world_normal_matrix; if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_NON_UNIFORM_SCALE)) { - world_normal_matrix = inverse(mat3(world_matrix)); + world_normal_matrix = transpose(inverse(mat3(world_matrix))); } else { world_normal_matrix = mat3(world_matrix); } @@ -356,6 +356,27 @@ void main() { #VERSION_DEFINES +/* Specialization Constants (Toggles) */ + +layout(constant_id = 0) const bool sc_use_forward_gi = false; +layout(constant_id = 1) const bool sc_use_light_projector = false; +layout(constant_id = 2) const bool sc_use_light_soft_shadows = false; +layout(constant_id = 3) const bool sc_use_directional_soft_shadows = false; + +/* Specialization Constants (Values) */ + +layout(constant_id = 6) const uint sc_soft_shadow_samples = 4; +layout(constant_id = 7) const uint sc_penumbra_shadow_samples = 4; + +layout(constant_id = 8) const uint sc_directional_soft_shadow_samples = 4; +layout(constant_id = 9) const uint sc_directional_penumbra_shadow_samples = 4; + +layout(constant_id = 10) const bool sc_decal_use_mipmaps = true; +layout(constant_id = 11) const bool sc_projector_use_mipmaps = true; + +// not used in clustered renderer but we share some code with the mobile renderer that requires this. +const float sc_luminance_multiplier = 1.0; + #include "scene_forward_clustered_inc.glsl" /* Varyings */ @@ -448,14 +469,15 @@ layout(location = 0) out vec4 frag_color; #if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) -#include "scene_forward_lights_inc.glsl" +/* Make a default specular mode SPECULAR_SCHLICK_GGX. */ +#if !defined(SPECULAR_DISABLED) && !defined(SPECULAR_SCHLICK_GGX) && !defined(SPECULAR_BLINN) && !defined(SPECULAR_PHONG) && !defined(SPECULAR_TOON) +#define SPECULAR_SCHLICK_GGX +#endif -#ifdef USE_FORWARD_GI +#include "scene_forward_lights_inc.glsl" #include "scene_forward_gi_inc.glsl" -#endif //USE_FORWARD_GI - #endif //!defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) #ifndef MODE_RENDER_DEPTH @@ -547,9 +569,8 @@ void main() { vec3 view = -normalize(vertex_interp); vec3 albedo = vec3(1.0); vec3 backlight = vec3(0.0); - vec4 transmittance_color = vec4(0.0); + vec4 transmittance_color = vec4(0.0, 0.0, 0.0, 1.0); float transmittance_depth = 0.0; - float transmittance_curve = 1.0; float transmittance_boost = 0.0; float metallic = 0.0; float specular = 0.5; @@ -634,12 +655,8 @@ void main() { } #ifdef LIGHT_TRANSMITTANCE_USED -#ifdef SSS_MODE_SKIN - transmittance_color.a = sss_strength; -#else transmittance_color.a *= sss_strength; #endif -#endif #ifndef USE_SHADOW_TO_OPACITY @@ -798,25 +815,35 @@ void main() { continue; //out of decal } - //we need ddx/ddy for mipmaps, so simulate them - vec2 ddx = (decals.data[decal_index].xform * vec4(vertex_ddx, 0.0)).xz; - vec2 ddy = (decals.data[decal_index].xform * vec4(vertex_ddy, 0.0)).xz; - float fade = pow(1.0 - (uv_local.y > 0.0 ? uv_local.y : -uv_local.y), uv_local.y > 0.0 ? decals.data[decal_index].upper_fade : decals.data[decal_index].lower_fade); if (decals.data[decal_index].normal_fade > 0.0) { fade *= smoothstep(decals.data[decal_index].normal_fade, 1.0, dot(normal_interp, decals.data[decal_index].normal) * 0.5 + 0.5); } + //we need ddx/ddy for mipmaps, so simulate them + vec2 ddx = (decals.data[decal_index].xform * vec4(vertex_ddx, 0.0)).xz; + vec2 ddy = (decals.data[decal_index].xform * vec4(vertex_ddy, 0.0)).xz; + if (decals.data[decal_index].albedo_rect != vec4(0.0)) { //has albedo - vec4 decal_albedo = textureGrad(sampler2D(decal_atlas_srgb, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uv_local.xz * decals.data[decal_index].albedo_rect.zw + decals.data[decal_index].albedo_rect.xy, ddx * decals.data[decal_index].albedo_rect.zw, ddy * decals.data[decal_index].albedo_rect.zw); + vec4 decal_albedo; + if (sc_decal_use_mipmaps) { + decal_albedo = textureGrad(sampler2D(decal_atlas_srgb, decal_sampler), uv_local.xz * decals.data[decal_index].albedo_rect.zw + decals.data[decal_index].albedo_rect.xy, ddx * decals.data[decal_index].albedo_rect.zw, ddy * decals.data[decal_index].albedo_rect.zw); + } else { + decal_albedo = textureLod(sampler2D(decal_atlas_srgb, decal_sampler), uv_local.xz * decals.data[decal_index].albedo_rect.zw + decals.data[decal_index].albedo_rect.xy, 0.0); + } decal_albedo *= decals.data[decal_index].modulate; decal_albedo.a *= fade; albedo = mix(albedo, decal_albedo.rgb, decal_albedo.a * decals.data[decal_index].albedo_mix); if (decals.data[decal_index].normal_rect != vec4(0.0)) { - vec3 decal_normal = textureGrad(sampler2D(decal_atlas, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uv_local.xz * decals.data[decal_index].normal_rect.zw + decals.data[decal_index].normal_rect.xy, ddx * decals.data[decal_index].normal_rect.zw, ddy * decals.data[decal_index].normal_rect.zw).xyz; + vec3 decal_normal; + if (sc_decal_use_mipmaps) { + decal_normal = textureGrad(sampler2D(decal_atlas, decal_sampler), uv_local.xz * decals.data[decal_index].normal_rect.zw + decals.data[decal_index].normal_rect.xy, ddx * decals.data[decal_index].normal_rect.zw, ddy * decals.data[decal_index].normal_rect.zw).xyz; + } else { + decal_normal = textureLod(sampler2D(decal_atlas, decal_sampler), uv_local.xz * decals.data[decal_index].normal_rect.zw + decals.data[decal_index].normal_rect.xy, 0.0).xyz; + } decal_normal.xy = decal_normal.xy * vec2(2.0, -2.0) - vec2(1.0, -1.0); //users prefer flipped y normal maps in most authoring software decal_normal.z = sqrt(max(0.0, 1.0 - dot(decal_normal.xy, decal_normal.xy))); //convert to view space, use xzy because y is up @@ -826,7 +853,12 @@ void main() { } if (decals.data[decal_index].orm_rect != vec4(0.0)) { - vec3 decal_orm = textureGrad(sampler2D(decal_atlas, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uv_local.xz * decals.data[decal_index].orm_rect.zw + decals.data[decal_index].orm_rect.xy, ddx * decals.data[decal_index].orm_rect.zw, ddy * decals.data[decal_index].orm_rect.zw).xyz; + vec3 decal_orm; + if (sc_decal_use_mipmaps) { + decal_orm = textureGrad(sampler2D(decal_atlas, decal_sampler), uv_local.xz * decals.data[decal_index].orm_rect.zw + decals.data[decal_index].orm_rect.xy, ddx * decals.data[decal_index].orm_rect.zw, ddy * decals.data[decal_index].orm_rect.zw).xyz; + } else { + decal_orm = textureLod(sampler2D(decal_atlas, decal_sampler), uv_local.xz * decals.data[decal_index].orm_rect.zw + decals.data[decal_index].orm_rect.xy, 0.0).xyz; + } ao = mix(ao, decal_orm.r, decal_albedo.a); roughness = mix(roughness, decal_orm.g, decal_albedo.a); metallic = mix(metallic, decal_orm.b, decal_albedo.a); @@ -835,7 +867,11 @@ void main() { if (decals.data[decal_index].emission_rect != vec4(0.0)) { //emission is additive, so its independent from albedo - emission += textureGrad(sampler2D(decal_atlas_srgb, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uv_local.xz * decals.data[decal_index].emission_rect.zw + decals.data[decal_index].emission_rect.xy, ddx * decals.data[decal_index].emission_rect.zw, ddy * decals.data[decal_index].emission_rect.zw).xyz * decals.data[decal_index].emission_energy * fade; + if (sc_decal_use_mipmaps) { + emission += textureGrad(sampler2D(decal_atlas_srgb, decal_sampler), uv_local.xz * decals.data[decal_index].emission_rect.zw + decals.data[decal_index].emission_rect.xy, ddx * decals.data[decal_index].emission_rect.zw, ddy * decals.data[decal_index].emission_rect.zw).xyz * decals.data[decal_index].emission_energy * fade; + } else { + emission += textureLod(sampler2D(decal_atlas_srgb, decal_sampler), uv_local.xz * decals.data[decal_index].emission_rect.zw + decals.data[decal_index].emission_rect.xy, 0.0).xyz * decals.data[decal_index].emission_energy * fade; + } } } } @@ -848,7 +884,7 @@ void main() { #ifdef NORMAL_USED if (scene_data.roughness_limiter_enabled) { - //http://www.jp.square-enix.com/tech/library/pdf/ImprovedGeometricSpecularAA.pdf + //https://www.jp.square-enix.com/tech/library/pdf/ImprovedGeometricSpecularAA.pdf float roughness2 = roughness * roughness; vec3 dndu = dFdx(normal), dndv = dFdy(normal); float variance = scene_data.roughness_limiter_amount * (dot(dndu, dndu) + dot(dndv, dndv)); @@ -879,6 +915,8 @@ void main() { specular_light = textureLod(samplerCube(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), ref_vec, roughness * MAX_ROUGHNESS_LOD).rgb; #endif //USE_RADIANCE_CUBEMAP_ARRAY + float horizon = min(1.0 + dot(ref_vec, normal), 1.0); + specular_light *= horizon * horizon; specular_light *= scene_data.ambient_light_color_energy.a; } @@ -904,7 +942,7 @@ void main() { } #endif // USE_LIGHTMAP #if defined(CUSTOM_IRRADIANCE_USED) - ambient_light = mix(specular_light, custom_irradiance.rgb, custom_irradiance.a); + ambient_light = mix(ambient_light, custom_irradiance.rgb, custom_irradiance.a); #endif #endif //!defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) @@ -968,9 +1006,9 @@ void main() { ambient_light += textureLod(sampler2DArray(lightmap_textures[ofs], material_samplers[SAMPLER_LINEAR_CLAMP]), uvw, 0.0).rgb; } } -#elif defined(USE_FORWARD_GI) +#else - if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_SDFGI)) { //has lightmap capture + if (sc_use_forward_gi && bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_SDFGI)) { //has lightmap capture //make vertex orientation the world one, but still align to camera vec3 cam_pos = mat3(scene_data.camera_matrix) * vertex; @@ -1042,7 +1080,7 @@ void main() { } } - if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_VOXEL_GI)) { // process voxel_gi_instances + if (sc_use_forward_gi && bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_VOXEL_GI)) { // process voxel_gi_instances uint index1 = instances.data[instance_index].gi_offset & 0xFFFF; vec3 ref_vec = normalize(reflect(normalize(vertex), normal)); @@ -1073,9 +1111,8 @@ void main() { specular_light = spec_accum.rgb; ambient_light = amb_accum.rgb; } -#else - if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_GI_BUFFERS)) { //use GI buffers + if (!sc_use_forward_gi && bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_GI_BUFFERS)) { //use GI buffers vec2 coord; @@ -1106,7 +1143,7 @@ void main() { ambient_light = mix(ambient_light, buffer_ambient.rgb, buffer_ambient.a); specular_light = mix(specular_light, buffer_reflection.rgb, buffer_reflection.a); } -#endif +#endif // !USE_LIGHTMAP if (scene_data.ssao_enabled) { float ssao = texture(sampler2D(ao_buffer, material_samplers[SAMPLER_LINEAR_CLAMP]), screen_uv).r; @@ -1187,7 +1224,7 @@ void main() { specular_light *= specular * metallic * albedo * 2.0; #else - // scales the specular reflections, needs to be be computed before lighting happens, + // scales the specular reflections, needs to be computed before lighting happens, // but after environment, GI, and reflection probes are added // Environment brdf approximation (Lazarov 2013) // see https://www.unrealengine.com/en-US/blog/physically-based-shading-on-mobile @@ -1233,14 +1270,13 @@ void main() { float shadow = 1.0; -#ifdef USE_SOFT_SHADOWS //version with soft shadows, more expensive if (directional_lights.data[i].shadow_enabled) { - float depth_z = -vertex.z; + if (sc_use_directional_soft_shadows && directional_lights.data[i].softshadow_angle > 0) { + float depth_z = -vertex.z; - vec4 pssm_coord; - vec3 shadow_color = vec3(0.0); - vec3 light_dir = directional_lights.data[i].direction; + vec3 shadow_color = vec3(0.0); + vec3 light_dir = directional_lights.data[i].direction; #define BIAS_FUNC(m_var, m_idx) \ m_var.xyz += light_dir * directional_lights.data[i].shadow_bias[m_idx]; \ @@ -1248,168 +1284,105 @@ void main() { normal_bias -= light_dir * dot(light_dir, normal_bias); \ m_var.xyz += normal_bias; - if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { - vec4 v = vec4(vertex, 1.0); + uint blend_index = 0; - BIAS_FUNC(v, 0) + if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { + vec4 v = vec4(vertex, 1.0); - pssm_coord = (directional_lights.data[i].shadow_matrix1 * v); - pssm_coord /= pssm_coord.w; + BIAS_FUNC(v, 0) + + vec4 pssm_coord = (directional_lights.data[i].shadow_matrix1 * v); + pssm_coord /= pssm_coord.w; - if (directional_lights.data[i].softshadow_angle > 0) { float range_pos = dot(directional_lights.data[i].direction, v.xyz); float range_begin = directional_lights.data[i].shadow_range_begin.x; float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; vec2 tex_scale = directional_lights.data[i].uv_scale1 * test_radius; shadow = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); - } else { - shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + blend_index++; } - shadow_color = directional_lights.data[i].shadow_color1.rgb; - - } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { - vec4 v = vec4(vertex, 1.0); + if (blend_index < 2 && depth_z < directional_lights.data[i].shadow_split_offsets.y) { + vec4 v = vec4(vertex, 1.0); - BIAS_FUNC(v, 1) + BIAS_FUNC(v, 1) - pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); - pssm_coord /= pssm_coord.w; + vec4 pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); + pssm_coord /= pssm_coord.w; - if (directional_lights.data[i].softshadow_angle > 0) { float range_pos = dot(directional_lights.data[i].direction, v.xyz); float range_begin = directional_lights.data[i].shadow_range_begin.y; float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; vec2 tex_scale = directional_lights.data[i].uv_scale2 * test_radius; - shadow = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); - } else { - shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + float s = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + + if (blend_index == 0) { + shadow = s; + } else { + //blend + float blend = smoothstep(0.0, directional_lights.data[i].shadow_split_offsets.x, depth_z); + shadow = mix(shadow, s, blend); + } + + blend_index++; } - shadow_color = directional_lights.data[i].shadow_color2.rgb; - } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { - vec4 v = vec4(vertex, 1.0); + if (blend_index < 2 && depth_z < directional_lights.data[i].shadow_split_offsets.z) { + vec4 v = vec4(vertex, 1.0); - BIAS_FUNC(v, 2) + BIAS_FUNC(v, 2) - pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); - pssm_coord /= pssm_coord.w; + vec4 pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); + pssm_coord /= pssm_coord.w; - if (directional_lights.data[i].softshadow_angle > 0) { float range_pos = dot(directional_lights.data[i].direction, v.xyz); float range_begin = directional_lights.data[i].shadow_range_begin.z; float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; vec2 tex_scale = directional_lights.data[i].uv_scale3 * test_radius; - shadow = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); - } else { - shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); - } - - shadow_color = directional_lights.data[i].shadow_color3.rgb; - - } else { - vec4 v = vec4(vertex, 1.0); + float s = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); - BIAS_FUNC(v, 3) - - pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); - pssm_coord /= pssm_coord.w; + if (blend_index == 0) { + shadow = s; + } else { + //blend + float blend = smoothstep(directional_lights.data[i].shadow_split_offsets.x, directional_lights.data[i].shadow_split_offsets.y, depth_z); + shadow = mix(shadow, s, blend); + } - if (directional_lights.data[i].softshadow_angle > 0) { - float range_pos = dot(directional_lights.data[i].direction, v.xyz); - float range_begin = directional_lights.data[i].shadow_range_begin.w; - float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; - vec2 tex_scale = directional_lights.data[i].uv_scale4 * test_radius; - shadow = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); - } else { - shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + blend_index++; } - shadow_color = directional_lights.data[i].shadow_color4.rgb; - } - - if (directional_lights.data[i].blend_splits) { - vec3 shadow_color_blend = vec3(0.0); - float pssm_blend; - float shadow2; - - if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { + if (blend_index < 2) { vec4 v = vec4(vertex, 1.0); - BIAS_FUNC(v, 1) - pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); - pssm_coord /= pssm_coord.w; - if (directional_lights.data[i].softshadow_angle > 0) { - float range_pos = dot(directional_lights.data[i].direction, v.xyz); - float range_begin = directional_lights.data[i].shadow_range_begin.y; - float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; - vec2 tex_scale = directional_lights.data[i].uv_scale2 * test_radius; - shadow2 = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); - } else { - shadow2 = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); - } + BIAS_FUNC(v, 3) - pssm_blend = smoothstep(0.0, directional_lights.data[i].shadow_split_offsets.x, depth_z); - shadow_color_blend = directional_lights.data[i].shadow_color2.rgb; - } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { - vec4 v = vec4(vertex, 1.0); - BIAS_FUNC(v, 2) - pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); + vec4 pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); pssm_coord /= pssm_coord.w; - if (directional_lights.data[i].softshadow_angle > 0) { - float range_pos = dot(directional_lights.data[i].direction, v.xyz); - float range_begin = directional_lights.data[i].shadow_range_begin.z; - float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; - vec2 tex_scale = directional_lights.data[i].uv_scale3 * test_radius; - shadow2 = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); - } else { - shadow2 = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); - } - - pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.x, directional_lights.data[i].shadow_split_offsets.y, depth_z); + float range_pos = dot(directional_lights.data[i].direction, v.xyz); + float range_begin = directional_lights.data[i].shadow_range_begin.w; + float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; + vec2 tex_scale = directional_lights.data[i].uv_scale4 * test_radius; + float s = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); - shadow_color_blend = directional_lights.data[i].shadow_color3.rgb; - } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { - vec4 v = vec4(vertex, 1.0); - BIAS_FUNC(v, 3) - pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); - pssm_coord /= pssm_coord.w; - if (directional_lights.data[i].softshadow_angle > 0) { - float range_pos = dot(directional_lights.data[i].direction, v.xyz); - float range_begin = directional_lights.data[i].shadow_range_begin.w; - float test_radius = (range_pos - range_begin) * directional_lights.data[i].softshadow_angle; - vec2 tex_scale = directional_lights.data[i].uv_scale4 * test_radius; - shadow2 = sample_directional_soft_shadow(directional_shadow_atlas, pssm_coord.xyz, tex_scale * directional_lights.data[i].soft_shadow_scale); + if (blend_index == 0) { + shadow = s; } else { - shadow2 = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + //blend + float blend = smoothstep(directional_lights.data[i].shadow_split_offsets.y, directional_lights.data[i].shadow_split_offsets.z, depth_z); + shadow = mix(shadow, s, blend); } - - pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.y, directional_lights.data[i].shadow_split_offsets.z, depth_z); - shadow_color_blend = directional_lights.data[i].shadow_color4.rgb; - } else { - pssm_blend = 0.0; //if no blend, same coord will be used (divide by z will result in same value, and already cached) } - pssm_blend = sqrt(pssm_blend); - - shadow = mix(shadow, shadow2, pssm_blend); - shadow_color = mix(shadow_color, shadow_color_blend, pssm_blend); - } - - shadow = mix(shadow, 1.0, smoothstep(directional_lights.data[i].fade_from, directional_lights.data[i].fade_to, vertex.z)); //done with negative values for performance - #undef BIAS_FUNC - } -#else - // Soft shadow disabled version + } else { //no soft shadows - if (directional_lights.data[i].shadow_enabled) { - float depth_z = -vertex.z; + float depth_z = -vertex.z; - vec4 pssm_coord; - vec3 light_dir = directional_lights.data[i].direction; - vec3 base_normal_bias = normalize(normal_interp) * (1.0 - max(0.0, dot(light_dir, -normalize(normal_interp)))); + vec4 pssm_coord; + vec3 light_dir = directional_lights.data[i].direction; + vec3 base_normal_bias = normalize(normal_interp) * (1.0 - max(0.0, dot(light_dir, -normalize(normal_interp)))); #define BIAS_FUNC(m_var, m_idx) \ m_var.xyz += light_dir * directional_lights.data[i].shadow_bias[m_idx]; \ @@ -1417,122 +1390,70 @@ void main() { normal_bias -= light_dir * dot(light_dir, normal_bias); \ m_var.xyz += normal_bias; - if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { - vec4 v = vec4(vertex, 1.0); - - BIAS_FUNC(v, 0) - - pssm_coord = (directional_lights.data[i].shadow_matrix1 * v); -#ifdef LIGHT_TRANSMITTANCE_USED - { - vec4 trans_vertex = vec4(vertex - normalize(normal_interp) * directional_lights.data[i].shadow_transmittance_bias.x, 1.0); - vec4 trans_coord = directional_lights.data[i].shadow_matrix1 * trans_vertex; - trans_coord /= trans_coord.w; - - float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; - shadow_z *= directional_lights.data[i].shadow_transmittance_z_scale.x; - float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.x; - - transmittance_z = z - shadow_z; - } -#endif - } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { - vec4 v = vec4(vertex, 1.0); - - BIAS_FUNC(v, 1) - - pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); -#ifdef LIGHT_TRANSMITTANCE_USED - { - vec4 trans_vertex = vec4(vertex - normalize(normal_interp) * directional_lights.data[i].shadow_transmittance_bias.y, 1.0); - vec4 trans_coord = directional_lights.data[i].shadow_matrix2 * trans_vertex; - trans_coord /= trans_coord.w; - - float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; - shadow_z *= directional_lights.data[i].shadow_transmittance_z_scale.y; - float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.y; - - transmittance_z = z - shadow_z; - } -#endif - } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { - vec4 v = vec4(vertex, 1.0); - - BIAS_FUNC(v, 2) - - pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); -#ifdef LIGHT_TRANSMITTANCE_USED - { - vec4 trans_vertex = vec4(vertex - normalize(normal_interp) * directional_lights.data[i].shadow_transmittance_bias.z, 1.0); - vec4 trans_coord = directional_lights.data[i].shadow_matrix3 * trans_vertex; - trans_coord /= trans_coord.w; - - float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; - shadow_z *= directional_lights.data[i].shadow_transmittance_z_scale.z; - float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.z; - - transmittance_z = z - shadow_z; - } -#endif - - } else { - vec4 v = vec4(vertex, 1.0); - - BIAS_FUNC(v, 3) - - pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); -#ifdef LIGHT_TRANSMITTANCE_USED - { - vec4 trans_vertex = vec4(vertex - normalize(normal_interp) * directional_lights.data[i].shadow_transmittance_bias.w, 1.0); - vec4 trans_coord = directional_lights.data[i].shadow_matrix4 * trans_vertex; - trans_coord /= trans_coord.w; - - float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; - shadow_z *= directional_lights.data[i].shadow_transmittance_z_scale.w; - float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.w; - - transmittance_z = z - shadow_z; - } -#endif - } - - pssm_coord /= pssm_coord.w; - - shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { + vec4 v = vec4(vertex, 1.0); - if (directional_lights.data[i].blend_splits) { - float pssm_blend; + BIAS_FUNC(v, 0) - if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { + pssm_coord = (directional_lights.data[i].shadow_matrix1 * v); + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 1) + pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); - pssm_blend = smoothstep(0.0, directional_lights.data[i].shadow_split_offsets.x, depth_z); - } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 2) + pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); - pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.x, directional_lights.data[i].shadow_split_offsets.y, depth_z); - } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { + + } else { vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 3) + pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); - pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.y, directional_lights.data[i].shadow_split_offsets.z, depth_z); - } else { - pssm_blend = 0.0; //if no blend, same coord will be used (divide by z will result in same value, and already cached) } pssm_coord /= pssm_coord.w; - float shadow2 = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); - shadow = mix(shadow, shadow2, pssm_blend); - } + shadow = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + + if (directional_lights.data[i].blend_splits) { + float pssm_blend; + + if (depth_z < directional_lights.data[i].shadow_split_offsets.x) { + vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 1) + pssm_coord = (directional_lights.data[i].shadow_matrix2 * v); + pssm_blend = smoothstep(0.0, directional_lights.data[i].shadow_split_offsets.x, depth_z); + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { + vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 2) + pssm_coord = (directional_lights.data[i].shadow_matrix3 * v); + pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.x, directional_lights.data[i].shadow_split_offsets.y, depth_z); + } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { + vec4 v = vec4(vertex, 1.0); + BIAS_FUNC(v, 3) + pssm_coord = (directional_lights.data[i].shadow_matrix4 * v); + pssm_blend = smoothstep(directional_lights.data[i].shadow_split_offsets.y, directional_lights.data[i].shadow_split_offsets.z, depth_z); + } else { + pssm_blend = 0.0; //if no blend, same coord will be used (divide by z will result in same value, and already cached) + } + + pssm_coord /= pssm_coord.w; + + float shadow2 = sample_directional_pcf_shadow(directional_shadow_atlas, scene_data.directional_shadow_pixel_size * directional_lights.data[i].soft_shadow_scale, pssm_coord); + shadow = mix(shadow, shadow2, pssm_blend); + } - shadow = mix(shadow, 1.0, smoothstep(directional_lights.data[i].fade_from, directional_lights.data[i].fade_to, vertex.z)); //done with negative values for performance + shadow = mix(shadow, 1.0, smoothstep(directional_lights.data[i].fade_from, directional_lights.data[i].fade_to, vertex.z)); //done with negative values for performance #undef BIAS_FUNC - } -#endif + } + } // shadows if (i < 4) { shadow0 |= uint(clamp(shadow * 255.0, 0.0, 255.0)) << (i * 8); @@ -1562,8 +1483,8 @@ void main() { trans_coord /= trans_coord.w; float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; - shadow_z *= directional_lights.data[i].shadow_transmittance_z_scale.x; - float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.x; + shadow_z *= directional_lights.data[i].shadow_z_range.x; + float z = trans_coord.z * directional_lights.data[i].shadow_z_range.x; transmittance_z = z - shadow_z; } else if (depth_z < directional_lights.data[i].shadow_split_offsets.y) { @@ -1572,8 +1493,8 @@ void main() { trans_coord /= trans_coord.w; float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; - shadow_z *= directional_lights.data[i].shadow_transmittance_z_scale.y; - float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.y; + shadow_z *= directional_lights.data[i].shadow_z_range.y; + float z = trans_coord.z * directional_lights.data[i].shadow_z_range.y; transmittance_z = z - shadow_z; } else if (depth_z < directional_lights.data[i].shadow_split_offsets.z) { @@ -1582,8 +1503,8 @@ void main() { trans_coord /= trans_coord.w; float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; - shadow_z *= directional_lights.data[i].shadow_transmittance_z_scale.z; - float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.z; + shadow_z *= directional_lights.data[i].shadow_z_range.z; + float z = trans_coord.z * directional_lights.data[i].shadow_z_range.z; transmittance_z = z - shadow_z; @@ -1593,221 +1514,218 @@ void main() { trans_coord /= trans_coord.w; float shadow_z = textureLod(sampler2D(directional_shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), trans_coord.xy, 0.0).r; - shadow_z *= directional_lights.data[i].shadow_transmittance_z_scale.w; - float z = trans_coord.z * directional_lights.data[i].shadow_transmittance_z_scale.w; + shadow_z *= directional_lights.data[i].shadow_z_range.w; + float z = trans_coord.z * directional_lights.data[i].shadow_z_range.w; transmittance_z = z - shadow_z; } + } #endif - float shadow = 1.0; + float shadow = 1.0; - if (i < 4) { - shadow = float(shadow0 >> (i * 8) & 0xFF) / 255.0; - } else { - shadow = float(shadow1 >> ((i - 4) * 8) & 0xFF) / 255.0; - } + if (i < 4) { + shadow = float(shadow0 >> (i * 8) & 0xFF) / 255.0; + } else { + shadow = float(shadow1 >> ((i - 4) * 8) & 0xFF) / 255.0; + } - blur_shadow(shadow); + blur_shadow(shadow); - light_compute(normal, directional_lights.data[i].direction, normalize(view), directional_lights.data[i].color * directional_lights.data[i].energy, shadow, f0, orms, 1.0, + float size_A = sc_use_light_soft_shadows ? directional_lights.data[i].size : 0.0; + + light_compute(normal, directional_lights.data[i].direction, normalize(view), size_A, directional_lights.data[i].color * directional_lights.data[i].energy, shadow, f0, orms, 1.0, #ifdef LIGHT_BACKLIGHT_USED - backlight, + backlight, #endif #ifdef LIGHT_TRANSMITTANCE_USED - transmittance_color, - transmittance_depth, - transmittance_curve, - transmittance_boost, - transmittance_z, + transmittance_color, + transmittance_depth, + transmittance_boost, + transmittance_z, #endif #ifdef LIGHT_RIM_USED - rim, rim_tint, albedo, + rim, rim_tint, albedo, #endif #ifdef LIGHT_CLEARCOAT_USED - clearcoat, clearcoat_gloss, + clearcoat, clearcoat_gloss, #endif #ifdef LIGHT_ANISOTROPY_USED - binormal, tangent, anisotropy, -#endif -#ifdef USE_SOFT_SHADOW - directional_lights.data[i].size, + binormal, tangent, anisotropy, #endif #ifdef USE_SHADOW_TO_OPACITY - alpha, + alpha, #endif - diffuse_light, - specular_light); - } + diffuse_light, + specular_light); } + } - { //omni lights + { //omni lights - uint cluster_omni_offset = cluster_offset; + uint cluster_omni_offset = cluster_offset; - uint item_min; - uint item_max; - uint item_from; - uint item_to; + uint item_min; + uint item_max; + uint item_from; + uint item_to; - cluster_get_item_range(cluster_omni_offset + scene_data.max_cluster_element_count_div_32 + cluster_z, item_min, item_max, item_from, item_to); + cluster_get_item_range(cluster_omni_offset + scene_data.max_cluster_element_count_div_32 + cluster_z, item_min, item_max, item_from, item_to); #ifdef USE_SUBGROUPS - item_from = subgroupBroadcastFirst(subgroupMin(item_from)); - item_to = subgroupBroadcastFirst(subgroupMax(item_to)); + item_from = subgroupBroadcastFirst(subgroupMin(item_from)); + item_to = subgroupBroadcastFirst(subgroupMax(item_to)); #endif - for (uint i = item_from; i < item_to; i++) { - uint mask = cluster_buffer.data[cluster_omni_offset + i]; - mask &= cluster_get_range_clip_mask(i, item_min, item_max); + for (uint i = item_from; i < item_to; i++) { + uint mask = cluster_buffer.data[cluster_omni_offset + i]; + mask &= cluster_get_range_clip_mask(i, item_min, item_max); #ifdef USE_SUBGROUPS - uint merged_mask = subgroupBroadcastFirst(subgroupOr(mask)); + uint merged_mask = subgroupBroadcastFirst(subgroupOr(mask)); #else uint merged_mask = mask; #endif - while (merged_mask != 0) { - uint bit = findMSB(merged_mask); - merged_mask &= ~(1 << bit); + while (merged_mask != 0) { + uint bit = findMSB(merged_mask); + merged_mask &= ~(1 << bit); #ifdef USE_SUBGROUPS - if (((1 << bit) & mask) == 0) { //do not process if not originally here - continue; - } + if (((1 << bit) & mask) == 0) { //do not process if not originally here + continue; + } #endif - uint light_index = 32 * i + bit; + uint light_index = 32 * i + bit; - if (!bool(omni_lights.data[light_index].mask & instances.data[instance_index].layer_mask)) { - continue; //not masked - } + if (!bool(omni_lights.data[light_index].mask & instances.data[instance_index].layer_mask)) { + continue; //not masked + } - if (omni_lights.data[light_index].bake_mode == LIGHT_BAKE_STATIC && bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_LIGHTMAP)) { - continue; // Statically baked light and object uses lightmap, skip - } + if (omni_lights.data[light_index].bake_mode == LIGHT_BAKE_STATIC && bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_LIGHTMAP)) { + continue; // Statically baked light and object uses lightmap, skip + } - float shadow = light_process_omni_shadow(light_index, vertex, view); + float shadow = light_process_omni_shadow(light_index, vertex, normal); - shadow = blur_shadow(shadow); + shadow = blur_shadow(shadow); - light_process_omni(light_index, vertex, view, normal, vertex_ddx, vertex_ddy, f0, orms, shadow, + light_process_omni(light_index, vertex, view, normal, vertex_ddx, vertex_ddy, f0, orms, shadow, #ifdef LIGHT_BACKLIGHT_USED - backlight, + backlight, #endif #ifdef LIGHT_TRANSMITTANCE_USED - transmittance_color, - transmittance_depth, - transmittance_curve, - transmittance_boost, + transmittance_color, + transmittance_depth, + transmittance_boost, #endif #ifdef LIGHT_RIM_USED - rim, - rim_tint, - albedo, + rim, + rim_tint, + albedo, #endif #ifdef LIGHT_CLEARCOAT_USED - clearcoat, clearcoat_gloss, + clearcoat, clearcoat_gloss, #endif #ifdef LIGHT_ANISOTROPY_USED - tangent, binormal, anisotropy, + tangent, binormal, anisotropy, #endif #ifdef USE_SHADOW_TO_OPACITY - alpha, + alpha, #endif - diffuse_light, specular_light); - } + diffuse_light, specular_light); } } + } - { //spot lights + { //spot lights - uint cluster_spot_offset = cluster_offset + scene_data.cluster_type_size; + uint cluster_spot_offset = cluster_offset + scene_data.cluster_type_size; - uint item_min; - uint item_max; - uint item_from; - uint item_to; + uint item_min; + uint item_max; + uint item_from; + uint item_to; - cluster_get_item_range(cluster_spot_offset + scene_data.max_cluster_element_count_div_32 + cluster_z, item_min, item_max, item_from, item_to); + cluster_get_item_range(cluster_spot_offset + scene_data.max_cluster_element_count_div_32 + cluster_z, item_min, item_max, item_from, item_to); #ifdef USE_SUBGROUPS - item_from = subgroupBroadcastFirst(subgroupMin(item_from)); - item_to = subgroupBroadcastFirst(subgroupMax(item_to)); + item_from = subgroupBroadcastFirst(subgroupMin(item_from)); + item_to = subgroupBroadcastFirst(subgroupMax(item_to)); #endif - for (uint i = item_from; i < item_to; i++) { - uint mask = cluster_buffer.data[cluster_spot_offset + i]; - mask &= cluster_get_range_clip_mask(i, item_min, item_max); + for (uint i = item_from; i < item_to; i++) { + uint mask = cluster_buffer.data[cluster_spot_offset + i]; + mask &= cluster_get_range_clip_mask(i, item_min, item_max); #ifdef USE_SUBGROUPS - uint merged_mask = subgroupBroadcastFirst(subgroupOr(mask)); + uint merged_mask = subgroupBroadcastFirst(subgroupOr(mask)); #else uint merged_mask = mask; #endif - while (merged_mask != 0) { - uint bit = findMSB(merged_mask); - merged_mask &= ~(1 << bit); + while (merged_mask != 0) { + uint bit = findMSB(merged_mask); + merged_mask &= ~(1 << bit); #ifdef USE_SUBGROUPS - if (((1 << bit) & mask) == 0) { //do not process if not originally here - continue; - } + if (((1 << bit) & mask) == 0) { //do not process if not originally here + continue; + } #endif - uint light_index = 32 * i + bit; + uint light_index = 32 * i + bit; - if (!bool(spot_lights.data[light_index].mask & instances.data[instance_index].layer_mask)) { - continue; //not masked - } + if (!bool(spot_lights.data[light_index].mask & instances.data[instance_index].layer_mask)) { + continue; //not masked + } - if (spot_lights.data[light_index].bake_mode == LIGHT_BAKE_STATIC && bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_LIGHTMAP)) { - continue; // Statically baked light and object uses lightmap, skip - } + if (spot_lights.data[light_index].bake_mode == LIGHT_BAKE_STATIC && bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_LIGHTMAP)) { + continue; // Statically baked light and object uses lightmap, skip + } - float shadow = light_process_spot_shadow(light_index, vertex, view); + float shadow = light_process_spot_shadow(light_index, vertex, normal); - shadow = blur_shadow(shadow); + shadow = blur_shadow(shadow); - light_process_spot(light_index, vertex, view, normal, vertex_ddx, vertex_ddy, f0, orms, shadow, + light_process_spot(light_index, vertex, view, normal, vertex_ddx, vertex_ddy, f0, orms, shadow, #ifdef LIGHT_BACKLIGHT_USED - backlight, + backlight, #endif #ifdef LIGHT_TRANSMITTANCE_USED - transmittance_color, - transmittance_depth, - transmittance_curve, - transmittance_boost, + transmittance_color, + transmittance_depth, + transmittance_boost, #endif #ifdef LIGHT_RIM_USED - rim, - rim_tint, - albedo, + rim, + rim_tint, + albedo, #endif #ifdef LIGHT_CLEARCOAT_USED - clearcoat, clearcoat_gloss, + clearcoat, clearcoat_gloss, #endif #ifdef LIGHT_ANISOTROPY_USED - tangent, binormal, anisotropy, + tangent, binormal, anisotropy, #endif #ifdef USE_SHADOW_TO_OPACITY - alpha, + alpha, #endif - diffuse_light, specular_light); - } + diffuse_light, specular_light); } } + } #ifdef USE_SHADOW_TO_OPACITY - alpha = min(alpha, clamp(length(ambient_light), 0.0, 1.0)); + alpha = min(alpha, clamp(length(ambient_light), 0.0, 1.0)); #if defined(ALPHA_SCISSOR_USED) - if (alpha < alpha_scissor) { - discard; - } + if (alpha < alpha_scissor) { + discard; + } #endif // ALPHA_SCISSOR_USED #ifdef USE_OPAQUE_PREPASS - if (alpha < opaque_prepass_threshold) { - discard; - } + if (alpha < opaque_prepass_threshold) { + discard; + } #endif // USE_OPAQUE_PREPASS @@ -1819,126 +1737,126 @@ void main() { #ifdef MODE_RENDER_SDF - { - vec3 local_pos = (scene_data.sdf_to_bounds * vec4(vertex, 1.0)).xyz; - ivec3 grid_pos = scene_data.sdf_offset + ivec3(local_pos * vec3(scene_data.sdf_size)); - - uint albedo16 = 0x1; //solid flag - albedo16 |= clamp(uint(albedo.r * 31.0), 0, 31) << 11; - albedo16 |= clamp(uint(albedo.g * 31.0), 0, 31) << 6; - albedo16 |= clamp(uint(albedo.b * 31.0), 0, 31) << 1; - - imageStore(albedo_volume_grid, grid_pos, uvec4(albedo16)); - - uint facing_bits = 0; - const vec3 aniso_dir[6] = vec3[]( - vec3(1, 0, 0), - vec3(0, 1, 0), - vec3(0, 0, 1), - vec3(-1, 0, 0), - vec3(0, -1, 0), - vec3(0, 0, -1)); - - vec3 cam_normal = mat3(scene_data.camera_matrix) * normalize(normal_interp); - - float closest_dist = -1e20; - - for (uint i = 0; i < 6; i++) { - float d = dot(cam_normal, aniso_dir[i]); - if (d > closest_dist) { - closest_dist = d; - facing_bits = (1 << i); - } + { + vec3 local_pos = (scene_data.sdf_to_bounds * vec4(vertex, 1.0)).xyz; + ivec3 grid_pos = scene_data.sdf_offset + ivec3(local_pos * vec3(scene_data.sdf_size)); + + uint albedo16 = 0x1; //solid flag + albedo16 |= clamp(uint(albedo.r * 31.0), 0, 31) << 11; + albedo16 |= clamp(uint(albedo.g * 31.0), 0, 31) << 6; + albedo16 |= clamp(uint(albedo.b * 31.0), 0, 31) << 1; + + imageStore(albedo_volume_grid, grid_pos, uvec4(albedo16)); + + uint facing_bits = 0; + const vec3 aniso_dir[6] = vec3[]( + vec3(1, 0, 0), + vec3(0, 1, 0), + vec3(0, 0, 1), + vec3(-1, 0, 0), + vec3(0, -1, 0), + vec3(0, 0, -1)); + + vec3 cam_normal = mat3(scene_data.camera_matrix) * normalize(normal_interp); + + float closest_dist = -1e20; + + for (uint i = 0; i < 6; i++) { + float d = dot(cam_normal, aniso_dir[i]); + if (d > closest_dist) { + closest_dist = d; + facing_bits = (1 << i); } + } - imageAtomicOr(geom_facing_grid, grid_pos, facing_bits); //store facing bits + imageAtomicOr(geom_facing_grid, grid_pos, facing_bits); //store facing bits - if (length(emission) > 0.001) { - float lumas[6]; - vec3 light_total = vec3(0); + if (length(emission) > 0.001) { + float lumas[6]; + vec3 light_total = vec3(0); - for (int i = 0; i < 6; i++) { - float strength = max(0.0, dot(cam_normal, aniso_dir[i])); - vec3 light = emission * strength; - light_total += light; - lumas[i] = max(light.r, max(light.g, light.b)); - } + for (int i = 0; i < 6; i++) { + float strength = max(0.0, dot(cam_normal, aniso_dir[i])); + vec3 light = emission * strength; + light_total += light; + lumas[i] = max(light.r, max(light.g, light.b)); + } - float luma_total = max(light_total.r, max(light_total.g, light_total.b)); + float luma_total = max(light_total.r, max(light_total.g, light_total.b)); - uint light_aniso = 0; + uint light_aniso = 0; - for (int i = 0; i < 6; i++) { - light_aniso |= min(31, uint((lumas[i] / luma_total) * 31.0)) << (i * 5); - } + for (int i = 0; i < 6; i++) { + light_aniso |= min(31, uint((lumas[i] / luma_total) * 31.0)) << (i * 5); + } - //compress to RGBE9995 to save space + //compress to RGBE9995 to save space - const float pow2to9 = 512.0f; - const float B = 15.0f; - const float N = 9.0f; - const float LN2 = 0.6931471805599453094172321215; + const float pow2to9 = 512.0f; + const float B = 15.0f; + const float N = 9.0f; + const float LN2 = 0.6931471805599453094172321215; - float cRed = clamp(light_total.r, 0.0, 65408.0); - float cGreen = clamp(light_total.g, 0.0, 65408.0); - float cBlue = clamp(light_total.b, 0.0, 65408.0); + float cRed = clamp(light_total.r, 0.0, 65408.0); + float cGreen = clamp(light_total.g, 0.0, 65408.0); + float cBlue = clamp(light_total.b, 0.0, 65408.0); - float cMax = max(cRed, max(cGreen, cBlue)); + float cMax = max(cRed, max(cGreen, cBlue)); - float expp = max(-B - 1.0f, floor(log(cMax) / LN2)) + 1.0f + B; + float expp = max(-B - 1.0f, floor(log(cMax) / LN2)) + 1.0f + B; - float sMax = floor((cMax / pow(2.0f, expp - B - N)) + 0.5f); + float sMax = floor((cMax / pow(2.0f, expp - B - N)) + 0.5f); - float exps = expp + 1.0f; + float exps = expp + 1.0f; - if (0.0 <= sMax && sMax < pow2to9) { - exps = expp; - } + if (0.0 <= sMax && sMax < pow2to9) { + exps = expp; + } - float sRed = floor((cRed / pow(2.0f, exps - B - N)) + 0.5f); - float sGreen = floor((cGreen / pow(2.0f, exps - B - N)) + 0.5f); - float sBlue = floor((cBlue / pow(2.0f, exps - B - N)) + 0.5f); - //store as 8985 to have 2 extra neighbour bits - uint light_rgbe = ((uint(sRed) & 0x1FF) >> 1) | ((uint(sGreen) & 0x1FF) << 8) | (((uint(sBlue) & 0x1FF) >> 1) << 17) | ((uint(exps) & 0x1F) << 25); + float sRed = floor((cRed / pow(2.0f, exps - B - N)) + 0.5f); + float sGreen = floor((cGreen / pow(2.0f, exps - B - N)) + 0.5f); + float sBlue = floor((cBlue / pow(2.0f, exps - B - N)) + 0.5f); + //store as 8985 to have 2 extra neighbour bits + uint light_rgbe = ((uint(sRed) & 0x1FF) >> 1) | ((uint(sGreen) & 0x1FF) << 8) | (((uint(sBlue) & 0x1FF) >> 1) << 17) | ((uint(exps) & 0x1F) << 25); - imageStore(emission_grid, grid_pos, uvec4(light_rgbe)); - imageStore(emission_aniso_grid, grid_pos, uvec4(light_aniso)); - } + imageStore(emission_grid, grid_pos, uvec4(light_rgbe)); + imageStore(emission_aniso_grid, grid_pos, uvec4(light_aniso)); } + } #endif #ifdef MODE_RENDER_MATERIAL - albedo_output_buffer.rgb = albedo; - albedo_output_buffer.a = alpha; + albedo_output_buffer.rgb = albedo; + albedo_output_buffer.a = alpha; - normal_output_buffer.rgb = normal * 0.5 + 0.5; - normal_output_buffer.a = 0.0; - depth_output_buffer.r = -vertex.z; + normal_output_buffer.rgb = normal * 0.5 + 0.5; + normal_output_buffer.a = 0.0; + depth_output_buffer.r = -vertex.z; - orm_output_buffer.r = ao; - orm_output_buffer.g = roughness; - orm_output_buffer.b = metallic; - orm_output_buffer.a = sss_strength; + orm_output_buffer.r = ao; + orm_output_buffer.g = roughness; + orm_output_buffer.b = metallic; + orm_output_buffer.a = sss_strength; - emission_output_buffer.rgb = emission; - emission_output_buffer.a = 0.0; + emission_output_buffer.rgb = emission; + emission_output_buffer.a = 0.0; #endif #ifdef MODE_RENDER_NORMAL_ROUGHNESS - normal_roughness_output_buffer = vec4(normal * 0.5 + 0.5, roughness); + normal_roughness_output_buffer = vec4(normal * 0.5 + 0.5, roughness); #ifdef MODE_RENDER_VOXEL_GI - if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_VOXEL_GI)) { // process voxel_gi_instances - uint index1 = instances.data[instance_index].gi_offset & 0xFFFF; - uint index2 = instances.data[instance_index].gi_offset >> 16; - voxel_gi_buffer.x = index1 & 0xFF; - voxel_gi_buffer.y = index2 & 0xFF; - } else { - voxel_gi_buffer.x = 0xFF; - voxel_gi_buffer.y = 0xFF; - } + if (bool(instances.data[instance_index].flags & INSTANCE_FLAGS_USE_VOXEL_GI)) { // process voxel_gi_instances + uint index1 = instances.data[instance_index].gi_offset & 0xFFFF; + uint index2 = instances.data[instance_index].gi_offset >> 16; + voxel_gi_buffer.x = index1 & 0xFF; + voxel_gi_buffer.y = index2 & 0xFF; + } else { + voxel_gi_buffer.x = 0xFF; + voxel_gi_buffer.y = 0xFF; + } #endif #endif //MODE_RENDER_NORMAL_ROUGHNESS @@ -1996,4 +1914,4 @@ void main() { #endif //MODE_MULTIPLE_RENDER_TARGETS #endif //MODE_RENDER_DEPTH - } +} diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_clustered_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_clustered_inc.glsl index e64e52623e..b53bf6a6d4 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_clustered_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_clustered_inc.glsl @@ -52,6 +52,11 @@ layout(set = 0, binding = 1) uniform sampler material_samplers[12]; layout(set = 0, binding = 2) uniform sampler shadow_sampler; +layout(set = 0, binding = 3) uniform sampler decal_sampler; + +layout(set = 0, binding = 4) uniform sampler light_projector_sampler; + +#define INSTANCE_FLAGS_NON_UNIFORM_SCALE (1 << 5) #define INSTANCE_FLAGS_USE_GI_BUFFERS (1 << 6) #define INSTANCE_FLAGS_USE_SDFGI (1 << 7) #define INSTANCE_FLAGS_USE_LIGHTMAP_CAPTURE (1 << 8) @@ -66,24 +71,22 @@ layout(set = 0, binding = 2) uniform sampler shadow_sampler; //3 bits of stride #define INSTANCE_FLAGS_PARTICLE_TRAIL_MASK 0xFF -#define INSTANCE_FLAGS_NON_UNIFORM_SCALE (1 << 24) - -layout(set = 0, binding = 3, std430) restrict readonly buffer OmniLights { +layout(set = 0, binding = 5, std430) restrict readonly buffer OmniLights { LightData data[]; } omni_lights; -layout(set = 0, binding = 4, std430) restrict readonly buffer SpotLights { +layout(set = 0, binding = 6, std430) restrict readonly buffer SpotLights { LightData data[]; } spot_lights; -layout(set = 0, binding = 5, std430) restrict readonly buffer ReflectionProbeData { +layout(set = 0, binding = 7, std430) restrict readonly buffer ReflectionProbeData { ReflectionData data[]; } reflections; -layout(set = 0, binding = 6, std140) uniform DirectionalLights { +layout(set = 0, binding = 8, std140) uniform DirectionalLights { DirectionalLightData data[MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS]; } directional_lights; @@ -95,7 +98,7 @@ struct Lightmap { mat3 normal_xform; }; -layout(set = 0, binding = 7, std140) restrict readonly buffer Lightmaps { +layout(set = 0, binding = 9, std140) restrict readonly buffer Lightmaps { Lightmap data[]; } lightmaps; @@ -104,20 +107,20 @@ struct LightmapCapture { vec4 sh[9]; }; -layout(set = 0, binding = 8, std140) restrict readonly buffer LightmapCaptures { +layout(set = 0, binding = 10, std140) restrict readonly buffer LightmapCaptures { LightmapCapture data[]; } lightmap_captures; -layout(set = 0, binding = 9) uniform texture2D decal_atlas; -layout(set = 0, binding = 10) uniform texture2D decal_atlas_srgb; +layout(set = 0, binding = 11) uniform texture2D decal_atlas; +layout(set = 0, binding = 12) uniform texture2D decal_atlas_srgb; -layout(set = 0, binding = 11, std430) restrict readonly buffer Decals { +layout(set = 0, binding = 13, std430) restrict readonly buffer Decals { DecalData data[]; } decals; -layout(set = 0, binding = 12, std430) restrict readonly buffer GlobalVariableData { +layout(set = 0, binding = 14, std430) restrict readonly buffer GlobalVariableData { vec4 data[]; } global_variables; @@ -129,7 +132,7 @@ struct SDFVoxelGICascadeData { float to_cell; // 1/bounds * grid_size }; -layout(set = 0, binding = 13, std140) uniform SDFGI { +layout(set = 0, binding = 15, std140) uniform SDFGI { vec3 grid_size; uint max_cascades; @@ -174,17 +177,12 @@ layout(set = 1, binding = 0, std140) uniform SceneData { uint cluster_type_size; uint max_cluster_element_count_div_32; - //use vec4s because std140 doesnt play nice with vec2s, z and w are wasted + // Use vec4s because std140 doesn't play nice with vec2s, z and w are wasted. vec4 directional_penumbra_shadow_kernel[32]; vec4 directional_soft_shadow_kernel[32]; vec4 penumbra_shadow_kernel[32]; vec4 soft_shadow_kernel[32]; - uint directional_penumbra_shadow_samples; - uint directional_soft_shadow_samples; - uint penumbra_shadow_samples; - uint soft_shadow_samples; - vec4 ambient_light_color_energy; float ambient_color_sky_mix; diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_lights_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_lights_inc.glsl index 709ea45b88..f3db4abe3b 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_lights_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_lights_inc.glsl @@ -73,14 +73,13 @@ vec3 F0(float metallic, float specular, vec3 albedo) { return mix(vec3(dielectric), albedo, vec3(metallic)); } -void light_compute(vec3 N, vec3 L, vec3 V, vec3 light_color, float attenuation, vec3 f0, uint orms, float specular_amount, +void light_compute(vec3 N, vec3 L, vec3 V, float A, vec3 light_color, float attenuation, vec3 f0, uint orms, float specular_amount, #ifdef LIGHT_BACKLIGHT_USED vec3 backlight, #endif #ifdef LIGHT_TRANSMITTANCE_USED vec4 transmittance_color, float transmittance_depth, - float transmittance_curve, float transmittance_boost, float transmittance_z, #endif @@ -93,9 +92,6 @@ void light_compute(vec3 N, vec3 L, vec3 V, vec3 light_color, float attenuation, #ifdef LIGHT_ANISOTROPY_USED vec3 B, vec3 T, float anisotropy, #endif -#ifdef USE_SOFT_SHADOWS - float A, -#endif #ifdef USE_SHADOW_TO_OPACITY inout float alpha, #endif @@ -112,11 +108,7 @@ void light_compute(vec3 N, vec3 L, vec3 V, vec3 light_color, float attenuation, #else -#ifdef USE_SOFT_SHADOWS float NdotL = min(A + dot(N, L), 1.0); -#else - float NdotL = dot(N, L); -#endif float cNdotL = max(NdotL, 0.0); // clamped NdotL float NdotV = dot(N, V); float cNdotV = max(NdotV, 0.0); @@ -126,19 +118,11 @@ void light_compute(vec3 N, vec3 L, vec3 V, vec3 light_color, float attenuation, #endif #if defined(SPECULAR_BLINN) || defined(SPECULAR_SCHLICK_GGX) || defined(LIGHT_CLEARCOAT_USED) -#ifdef USE_SOFT_SHADOWS float cNdotH = clamp(A + dot(N, H), 0.0, 1.0); -#else - float cNdotH = clamp(dot(N, H), 0.0, 1.0); -#endif #endif #if defined(DIFFUSE_BURLEY) || defined(SPECULAR_SCHLICK_GGX) || defined(LIGHT_CLEARCOAT_USED) -#ifdef USE_SOFT_SHADOWS float cLdotH = clamp(A + dot(L, H), 0.0, 1.0); -#else - float cLdotH = clamp(dot(L, H), 0.0, 1.0); -#endif #endif float metallic = unpackUnorm4x8(orms).z; @@ -189,9 +173,8 @@ void light_compute(vec3 N, vec3 L, vec3 V, vec3 light_color, float attenuation, #ifdef LIGHT_TRANSMITTANCE_USED -#ifdef SSS_MODE_SKIN - { +#ifdef SSS_MODE_SKIN float scale = 8.25 / transmittance_depth; float d = scale * abs(transmittance_z); float dd = -d * d; @@ -203,19 +186,15 @@ void light_compute(vec3 N, vec3 L, vec3 V, vec3 light_color, float attenuation, vec3(0.078, 0.0, 0.0) * exp(dd / 7.41); diffuse_light += profile * transmittance_color.a * light_color * clamp(transmittance_boost - NdotL, 0.0, 1.0) * (1.0 / M_PI); - } #else - if (transmittance_depth > 0.0) { - float fade = clamp(abs(transmittance_z / transmittance_depth), 0.0, 1.0); - - fade = pow(max(0.0, 1.0 - fade), transmittance_curve); - fade *= clamp(transmittance_boost - NdotL, 0.0, 1.0); - - diffuse_light += transmittance_color.rgb * light_color * (1.0 / M_PI) * transmittance_color.a * fade; + float scale = 8.25 / transmittance_depth; + float d = scale * abs(transmittance_z); + float dd = -d * d; + diffuse_light += exp(dd) * transmittance_color.rgb * transmittance_color.a * light_color * clamp(transmittance_boost - NdotL, 0.0, 1.0) * (1.0 / M_PI); +#endif } - -#endif //SSS_MODE_SKIN +#else #endif //LIGHT_TRANSMITTANCE_USED } @@ -229,26 +208,20 @@ void light_compute(vec3 N, vec3 L, vec3 V, vec3 light_color, float attenuation, //normalized blinn float shininess = exp2(15.0 * (1.0 - roughness) + 1.0) * 0.25; - float blinn = pow(cNdotH, shininess) * cNdotL; - blinn *= (shininess + 8.0) * (1.0 / (8.0 * M_PI)); - float intensity = blinn; + float blinn = pow(cNdotH, shininess); + blinn *= (shininess + 2.0) * (1.0 / (8.0 * M_PI)); - specular_light += light_color * intensity * attenuation * specular_amount; + specular_light += light_color * attenuation * specular_amount * blinn * f0 * unpackUnorm4x8(orms).w; #elif defined(SPECULAR_PHONG) vec3 R = normalize(-reflect(L, N)); -#ifdef USE_SOFT_SHADOWS float cRdotV = clamp(A + dot(R, V), 0.0, 1.0); -#else - float cRdotV = clamp(dot(R, V), 0.0, 1.0); -#endif float shininess = exp2(15.0 * (1.0 - roughness) + 1.0) * 0.25; float phong = pow(cRdotV, shininess); - phong *= (shininess + 8.0) * (1.0 / (8.0 * M_PI)); - float intensity = (phong) / max(4.0 * cNdotV * cNdotL, 0.75); + phong *= (shininess + 1.0) * (1.0 / (8.0 * M_PI)); - specular_light += light_color * intensity * attenuation * specular_amount; + specular_light += light_color * attenuation * specular_amount * phong * f0 * unpackUnorm4x8(orms).w; #elif defined(SPECULAR_TOON) @@ -306,7 +279,7 @@ void light_compute(vec3 N, vec3 L, vec3 V, vec3 light_color, float attenuation, } #ifdef USE_SHADOW_TO_OPACITY - alpha = min(alpha, clamp(1.0 - attenuation), 0.0, 1.0)); + alpha = min(alpha, clamp(1.0 - attenuation, 0.0, 1.0)); #endif #endif //defined(LIGHT_CODE_USED) @@ -315,7 +288,7 @@ void light_compute(vec3 N, vec3 L, vec3 V, vec3 light_color, float attenuation, #ifndef USE_NO_SHADOWS // Interleaved Gradient Noise -// http://www.iryoku.com/next-generation-post-processing-in-call-of-duty-advanced-warfare +// https://www.iryoku.com/next-generation-post-processing-in-call-of-duty-advanced-warfare float quick_hash(vec2 pos) { const vec3 magic = vec3(0.06711056f, 0.00583715f, 52.9829189f); return fract(magic.z * fract(dot(pos, magic.xy))); @@ -326,7 +299,7 @@ float sample_directional_pcf_shadow(texture2D shadow, vec2 shadow_pixel_size, ve float depth = coord.z; //if only one sample is taken, take it from the center - if (scene_data.directional_soft_shadow_samples == 1) { + if (sc_directional_soft_shadow_samples == 1) { return textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos, depth, 1.0)); } @@ -340,19 +313,19 @@ float sample_directional_pcf_shadow(texture2D shadow, vec2 shadow_pixel_size, ve float avg = 0.0; - for (uint i = 0; i < scene_data.directional_soft_shadow_samples; i++) { + for (uint i = 0; i < sc_directional_soft_shadow_samples; i++) { avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + shadow_pixel_size * (disk_rotation * scene_data.directional_soft_shadow_kernel[i].xy), depth, 1.0)); } - return avg * (1.0 / float(scene_data.directional_soft_shadow_samples)); + return avg * (1.0 / float(sc_directional_soft_shadow_samples)); } -float sample_pcf_shadow(texture2D shadow, vec2 shadow_pixel_size, vec4 coord) { +float sample_pcf_shadow(texture2D shadow, vec2 shadow_pixel_size, vec3 coord) { vec2 pos = coord.xy; float depth = coord.z; //if only one sample is taken, take it from the center - if (scene_data.soft_shadow_samples == 1) { + if (sc_soft_shadow_samples == 1) { return textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos, depth, 1.0)); } @@ -366,11 +339,54 @@ float sample_pcf_shadow(texture2D shadow, vec2 shadow_pixel_size, vec4 coord) { float avg = 0.0; - for (uint i = 0; i < scene_data.soft_shadow_samples; i++) { + for (uint i = 0; i < sc_soft_shadow_samples; i++) { avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos + shadow_pixel_size * (disk_rotation * scene_data.soft_shadow_kernel[i].xy), depth, 1.0)); } - return avg * (1.0 / float(scene_data.soft_shadow_samples)); + return avg * (1.0 / float(sc_soft_shadow_samples)); +} + +float sample_omni_pcf_shadow(texture2D shadow, float blur_scale, vec2 coord, vec4 uv_rect, vec2 flip_offset, float depth) { + //if only one sample is taken, take it from the center + if (sc_soft_shadow_samples == 1) { + vec2 pos = coord * 0.5 + 0.5; + pos = uv_rect.xy + pos * uv_rect.zw; + return textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(pos, depth, 1.0)); + } + + mat2 disk_rotation; + { + float r = quick_hash(gl_FragCoord.xy) * 2.0 * M_PI; + float sr = sin(r); + float cr = cos(r); + disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); + } + + float avg = 0.0; + vec2 offset_scale = blur_scale * 2.0 * scene_data.shadow_atlas_pixel_size / uv_rect.zw; + + for (uint i = 0; i < sc_soft_shadow_samples; i++) { + vec2 offset = offset_scale * (disk_rotation * scene_data.soft_shadow_kernel[i].xy); + vec2 sample_coord = coord + offset; + + float sample_coord_length_sqaured = dot(sample_coord, sample_coord); + bool do_flip = sample_coord_length_sqaured > 1.0; + + if (do_flip) { + float len = sqrt(sample_coord_length_sqaured); + sample_coord = sample_coord * (2.0 / len - 1.0); + } + + sample_coord = sample_coord * 0.5 + 0.5; + sample_coord = uv_rect.xy + sample_coord * uv_rect.zw; + + if (do_flip) { + sample_coord += flip_offset; + } + avg += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(sample_coord, depth, 1.0)); + } + + return avg * (1.0 / float(sc_soft_shadow_samples)); } float sample_directional_soft_shadow(texture2D shadow, vec3 pssm_coord, vec2 tex_scale) { @@ -386,7 +402,7 @@ float sample_directional_soft_shadow(texture2D shadow, vec3 pssm_coord, vec2 tex disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); } - for (uint i = 0; i < scene_data.directional_penumbra_shadow_samples; i++) { + for (uint i = 0; i < sc_directional_penumbra_shadow_samples; i++) { vec2 suv = pssm_coord.xy + (disk_rotation * scene_data.directional_penumbra_shadow_kernel[i].xy) * tex_scale; float d = textureLod(sampler2D(shadow, material_samplers[SAMPLER_LINEAR_CLAMP]), suv, 0.0).r; if (d < pssm_coord.z) { @@ -402,12 +418,12 @@ float sample_directional_soft_shadow(texture2D shadow, vec3 pssm_coord, vec2 tex tex_scale *= penumbra; float s = 0.0; - for (uint i = 0; i < scene_data.directional_penumbra_shadow_samples; i++) { + for (uint i = 0; i < sc_directional_penumbra_shadow_samples; i++) { vec2 suv = pssm_coord.xy + (disk_rotation * scene_data.directional_penumbra_shadow_kernel[i].xy) * tex_scale; s += textureProj(sampler2DShadow(shadow, shadow_sampler), vec4(suv, pssm_coord.z, 1.0)); } - return s / float(scene_data.directional_penumbra_shadow_samples); + return s / float(sc_directional_penumbra_shadow_samples); } else { //no blockers found, so no shadow @@ -430,26 +446,25 @@ float light_process_omni_shadow(uint idx, vec3 vertex, vec3 normal) { #ifndef USE_NO_SHADOWS if (omni_lights.data[idx].shadow_enabled) { // there is a shadowmap + vec2 texel_size = scene_data.shadow_atlas_pixel_size; + vec4 base_uv_rect = omni_lights.data[idx].atlas_rect; + base_uv_rect.xy += texel_size; + base_uv_rect.zw -= texel_size * 2.0; - vec3 light_rel_vec = omni_lights.data[idx].position - vertex; - float light_length = length(light_rel_vec); + // Omni lights use direction.xy to store to store the offset between the two paraboloid regions + vec2 flip_offset = omni_lights.data[idx].direction.xy; - vec4 v = vec4(vertex, 1.0); + vec3 local_vert = (omni_lights.data[idx].shadow_matrix * vec4(vertex, 1.0)).xyz; - vec4 splane = (omni_lights.data[idx].shadow_matrix * v); - float shadow_len = length(splane.xyz); //need to remember shadow len from here + float shadow_len = length(local_vert); //need to remember shadow len from here + vec3 shadow_dir = normalize(local_vert); - { - vec3 nofs = normal_interp * omni_lights.data[idx].shadow_normal_bias / omni_lights.data[idx].inv_radius; - nofs *= (1.0 - max(0.0, dot(normalize(light_rel_vec), normalize(normal_interp)))); - v.xyz += nofs; - splane = (omni_lights.data[idx].shadow_matrix * v); - } + vec3 local_normal = normalize(mat3(omni_lights.data[idx].shadow_matrix) * normal); + vec3 normal_bias = local_normal * omni_lights.data[idx].shadow_normal_bias * (1.0 - abs(dot(local_normal, shadow_dir))); float shadow; -#ifdef USE_SOFT_SHADOWS - if (omni_lights.data[idx].soft_shadow_size > 0.0) { + if (sc_use_light_soft_shadows && omni_lights.data[idx].soft_shadow_size > 0.0) { //soft shadow //find blocker @@ -465,30 +480,29 @@ float light_process_omni_shadow(uint idx, vec3 vertex, vec3 normal) { disk_rotation = mat2(vec2(cr, -sr), vec2(sr, cr)); } - vec3 normal = normalize(splane.xyz); - vec3 v0 = abs(normal.z) < 0.999 ? vec3(0.0, 0.0, 1.0) : vec3(0.0, 1.0, 0.0); - vec3 tangent = normalize(cross(v0, normal)); - vec3 bitangent = normalize(cross(tangent, normal)); + vec3 basis_normal = shadow_dir; + vec3 v0 = abs(basis_normal.z) < 0.999 ? vec3(0.0, 0.0, 1.0) : vec3(0.0, 1.0, 0.0); + vec3 tangent = normalize(cross(v0, basis_normal)); + vec3 bitangent = normalize(cross(tangent, basis_normal)); float z_norm = shadow_len * omni_lights.data[idx].inv_radius; tangent *= omni_lights.data[idx].soft_shadow_size * omni_lights.data[idx].soft_shadow_scale; bitangent *= omni_lights.data[idx].soft_shadow_size * omni_lights.data[idx].soft_shadow_scale; - for (uint i = 0; i < scene_data.penumbra_shadow_samples; i++) { + for (uint i = 0; i < sc_penumbra_shadow_samples; i++) { vec2 disk = disk_rotation * scene_data.penumbra_shadow_kernel[i].xy; - vec3 pos = splane.xyz + tangent * disk.x + bitangent * disk.y; + vec3 pos = local_vert + tangent * disk.x + bitangent * disk.y; pos = normalize(pos); - vec4 uv_rect = omni_lights.data[idx].atlas_rect; + + vec4 uv_rect = base_uv_rect; if (pos.z >= 0.0) { - pos.z += 1.0; - uv_rect.y += uv_rect.w; - } else { - pos.z = 1.0 - pos.z; + uv_rect.xy += flip_offset; } + pos.z = 1.0 + abs(pos.z); pos.xy /= pos.z; pos.xy = pos.xy * 0.5 + 0.5; @@ -511,20 +525,20 @@ float light_process_omni_shadow(uint idx, vec3 vertex, vec3 normal) { z_norm -= omni_lights.data[idx].inv_radius * omni_lights.data[idx].shadow_bias; shadow = 0.0; - for (uint i = 0; i < scene_data.penumbra_shadow_samples; i++) { + for (uint i = 0; i < sc_penumbra_shadow_samples; i++) { vec2 disk = disk_rotation * scene_data.penumbra_shadow_kernel[i].xy; - vec3 pos = splane.xyz + tangent * disk.x + bitangent * disk.y; + vec3 pos = local_vert + tangent * disk.x + bitangent * disk.y; pos = normalize(pos); - vec4 uv_rect = omni_lights.data[idx].atlas_rect; + pos = normalize(pos + normal_bias); + + vec4 uv_rect = base_uv_rect; if (pos.z >= 0.0) { - pos.z += 1.0; - uv_rect.y += uv_rect.w; - } else { - pos.z = 1.0 - pos.z; + uv_rect.xy += flip_offset; } + pos.z = 1.0 + abs(pos.z); pos.xy /= pos.z; pos.xy = pos.xy * 0.5 + 0.5; @@ -532,36 +546,27 @@ float light_process_omni_shadow(uint idx, vec3 vertex, vec3 normal) { shadow += textureProj(sampler2DShadow(shadow_atlas, shadow_sampler), vec4(pos.xy, z_norm, 1.0)); } - shadow /= float(scene_data.penumbra_shadow_samples); + shadow /= float(sc_penumbra_shadow_samples); } else { //no blockers found, so no shadow shadow = 1.0; } } else { -#endif - splane.xyz = normalize(splane.xyz); - vec4 clamp_rect = omni_lights.data[idx].atlas_rect; - - if (splane.z >= 0.0) { - splane.z += 1.0; + vec4 uv_rect = base_uv_rect; - clamp_rect.y += clamp_rect.w; - - } else { - splane.z = 1.0 - splane.z; + vec3 shadow_sample = normalize(shadow_dir + normal_bias); + if (shadow_sample.z >= 0.0) { + uv_rect.xy += flip_offset; + flip_offset *= -1.0; } - splane.xy /= splane.z; - - splane.xy = splane.xy * 0.5 + 0.5; - splane.z = (shadow_len - omni_lights.data[idx].shadow_bias) * omni_lights.data[idx].inv_radius; - splane.xy = clamp_rect.xy + splane.xy * clamp_rect.zw; - splane.w = 1.0; //needed? i think it should be 1 already - shadow = sample_pcf_shadow(shadow_atlas, omni_lights.data[idx].soft_shadow_scale * scene_data.shadow_atlas_pixel_size, splane); -#ifdef USE_SOFT_SHADOWS + shadow_sample.z = 1.0 + abs(shadow_sample.z); + vec2 pos = shadow_sample.xy / shadow_sample.z; + float depth = shadow_len - omni_lights.data[idx].shadow_bias; + depth *= omni_lights.data[idx].inv_radius; + shadow = sample_omni_pcf_shadow(shadow_atlas, omni_lights.data[idx].soft_shadow_scale / shadow_sample.z, pos, uv_rect, flip_offset, depth); } -#endif return shadow; } @@ -577,7 +582,6 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v #ifdef LIGHT_TRANSMITTANCE_USED vec4 transmittance_color, float transmittance_depth, - float transmittance_curve, float transmittance_boost, #endif #ifdef LIGHT_RIM_USED @@ -599,14 +603,12 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v float light_attenuation = omni_attenuation; vec3 color = omni_lights.data[idx].color; -#ifdef USE_SOFT_SHADOWS float size_A = 0.0; - if (omni_lights.data[idx].size > 0.0) { + if (sc_use_light_soft_shadows && omni_lights.data[idx].size > 0.0) { float t = omni_lights.data[idx].size / max(0.001, light_length); size_A = max(0.0, 1.0 - 1 / sqrt(1 + t * t)); } -#endif #ifdef LIGHT_TRANSMITTANCE_USED float transmittance_z = transmittance_depth; //no transmittance by default @@ -617,20 +619,22 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v //redo shadowmapping, but shrink the model a bit to avoid arctifacts vec4 splane = (omni_lights.data[idx].shadow_matrix * vec4(vertex - normalize(normal_interp) * omni_lights.data[idx].transmittance_bias, 1.0)); - shadow_len = length(splane.xyz); - splane = normalize(splane.xyz); + float shadow_len = length(splane.xyz); + splane.xyz = normalize(splane.xyz); if (splane.z >= 0.0) { splane.z += 1.0; - + clamp_rect.y += clamp_rect.w; } else { splane.z = 1.0 - splane.z; } splane.xy /= splane.z; + splane.xy = splane.xy * 0.5 + 0.5; splane.z = shadow_len * omni_lights.data[idx].inv_radius; splane.xy = clamp_rect.xy + splane.xy * clamp_rect.zw; + // splane.xy = clamp(splane.xy,clamp_rect.xy + scene_data.shadow_atlas_pixel_size,clamp_rect.xy + clamp_rect.zw - scene_data.shadow_atlas_pixel_size ); splane.w = 1.0; //needed? i think it should be 1 already float shadow_z = textureLod(sampler2D(shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), splane.xy, 0.0).r; @@ -638,73 +642,72 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v } #endif -#if 0 - - if (omni_lights.data[idx].projector_rect != vec4(0.0)) { + if (sc_use_light_projector && omni_lights.data[idx].projector_rect != vec4(0.0)) { vec3 local_v = (omni_lights.data[idx].shadow_matrix * vec4(vertex, 1.0)).xyz; local_v = normalize(local_v); vec4 atlas_rect = omni_lights.data[idx].projector_rect; if (local_v.z >= 0.0) { - local_v.z += 1.0; atlas_rect.y += atlas_rect.w; - - } else { - local_v.z = 1.0 - local_v.z; } + local_v.z = 1.0 + abs(local_v.z); + local_v.xy /= local_v.z; local_v.xy = local_v.xy * 0.5 + 0.5; vec2 proj_uv = local_v.xy * atlas_rect.zw; - vec2 proj_uv_ddx; - vec2 proj_uv_ddy; - { - vec3 local_v_ddx = (omni_lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddx, 1.0)).xyz; - local_v_ddx = normalize(local_v_ddx); + if (sc_projector_use_mipmaps) { + vec2 proj_uv_ddx; + vec2 proj_uv_ddy; + { + vec3 local_v_ddx = (omni_lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddx, 1.0)).xyz; + local_v_ddx = normalize(local_v_ddx); - if (local_v_ddx.z >= 0.0) { - local_v_ddx.z += 1.0; - } else { - local_v_ddx.z = 1.0 - local_v_ddx.z; - } + if (local_v_ddx.z >= 0.0) { + local_v_ddx.z += 1.0; + } else { + local_v_ddx.z = 1.0 - local_v_ddx.z; + } - local_v_ddx.xy /= local_v_ddx.z; - local_v_ddx.xy = local_v_ddx.xy * 0.5 + 0.5; + local_v_ddx.xy /= local_v_ddx.z; + local_v_ddx.xy = local_v_ddx.xy * 0.5 + 0.5; - proj_uv_ddx = local_v_ddx.xy * atlas_rect.zw - proj_uv; + proj_uv_ddx = local_v_ddx.xy * atlas_rect.zw - proj_uv; - vec3 local_v_ddy = (omni_lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddy, 1.0)).xyz; - local_v_ddy = normalize(local_v_ddy); + vec3 local_v_ddy = (omni_lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddy, 1.0)).xyz; + local_v_ddy = normalize(local_v_ddy); - if (local_v_ddy.z >= 0.0) { - local_v_ddy.z += 1.0; - } else { - local_v_ddy.z = 1.0 - local_v_ddy.z; - } + if (local_v_ddy.z >= 0.0) { + local_v_ddy.z += 1.0; + } else { + local_v_ddy.z = 1.0 - local_v_ddy.z; + } - local_v_ddy.xy /= local_v_ddy.z; - local_v_ddy.xy = local_v_ddy.xy * 0.5 + 0.5; + local_v_ddy.xy /= local_v_ddy.z; + local_v_ddy.xy = local_v_ddy.xy * 0.5 + 0.5; - proj_uv_ddy = local_v_ddy.xy * atlas_rect.zw - proj_uv; - } + proj_uv_ddy = local_v_ddy.xy * atlas_rect.zw - proj_uv; + } - vec4 proj = textureGrad(sampler2D(decal_atlas_srgb, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), proj_uv + atlas_rect.xy, proj_uv_ddx, proj_uv_ddy); - no_shadow = mix(no_shadow, proj.rgb, proj.a); + vec4 proj = textureGrad(sampler2D(decal_atlas_srgb, light_projector_sampler), proj_uv + atlas_rect.xy, proj_uv_ddx, proj_uv_ddy); + color *= proj.rgb * proj.a; + } else { + vec4 proj = textureLod(sampler2D(decal_atlas_srgb, light_projector_sampler), proj_uv + atlas_rect.xy, 0.0); + color *= proj.rgb * proj.a; + } } -#endif light_attenuation *= shadow; - light_compute(normal, normalize(light_rel_vec), eye_vec, color, light_attenuation, f0, orms, omni_lights.data[idx].specular_amount, + light_compute(normal, normalize(light_rel_vec), eye_vec, size_A, color, light_attenuation, f0, orms, omni_lights.data[idx].specular_amount, #ifdef LIGHT_BACKLIGHT_USED backlight, #endif #ifdef LIGHT_TRANSMITTANCE_USED transmittance_color, transmittance_depth, - transmittance_curve, transmittance_boost, transmittance_z, #endif @@ -717,9 +720,6 @@ void light_process_omni(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v #ifdef LIGHT_ANISOTROPY_USED binormal, tangent, anisotropy, #endif -#ifdef USE_SOFT_SHADOWS - size_A, -#endif #ifdef USE_SHADOW_TO_OPACITY alpha, #endif @@ -733,31 +733,23 @@ float light_process_spot_shadow(uint idx, vec3 vertex, vec3 normal) { vec3 light_rel_vec = spot_lights.data[idx].position - vertex; float light_length = length(light_rel_vec); vec3 spot_dir = spot_lights.data[idx].direction; - //there is a shadowmap - vec4 v = vec4(vertex, 1.0); - - v.xyz -= spot_dir * spot_lights.data[idx].shadow_bias; - - float z_norm = dot(spot_dir, -light_rel_vec) * spot_lights.data[idx].inv_radius; - float depth_bias_scale = 1.0 / (max(0.0001, z_norm)); //the closer to the light origin, the more you have to offset to reach 1px in the map - vec3 normal_bias = normalize(normal_interp) * (1.0 - max(0.0, dot(spot_dir, -normalize(normal_interp)))) * spot_lights.data[idx].shadow_normal_bias * depth_bias_scale; - normal_bias -= spot_dir * dot(spot_dir, normal_bias); //only XY, no Z - v.xyz += normal_bias; + vec3 shadow_dir = light_rel_vec / light_length; + vec3 normal_bias = normal * light_length * spot_lights.data[idx].shadow_normal_bias * (1.0 - abs(dot(normal, shadow_dir))); - //adjust with bias - z_norm = dot(spot_dir, v.xyz - spot_lights.data[idx].position) * spot_lights.data[idx].inv_radius; - - float shadow; + //there is a shadowmap + vec4 v = vec4(vertex + normal_bias, 1.0); vec4 splane = (spot_lights.data[idx].shadow_matrix * v); + splane.z -= spot_lights.data[idx].shadow_bias / (light_length * spot_lights.data[idx].inv_radius); splane /= splane.w; -#ifdef USE_SOFT_SHADOWS - if (spot_lights.data[idx].soft_shadow_size > 0.0) { + float shadow; + if (sc_use_light_soft_shadows && spot_lights.data[idx].soft_shadow_size > 0.0) { //soft shadow //find blocker + float z_norm = dot(spot_dir, -light_rel_vec) * spot_lights.data[idx].inv_radius; vec2 shadow_uv = splane.xy * spot_lights.data[idx].atlas_rect.zw + spot_lights.data[idx].atlas_rect.xy; @@ -774,11 +766,11 @@ float light_process_spot_shadow(uint idx, vec3 vertex, vec3 normal) { float uv_size = spot_lights.data[idx].soft_shadow_size * z_norm * spot_lights.data[idx].soft_shadow_scale; vec2 clamp_max = spot_lights.data[idx].atlas_rect.xy + spot_lights.data[idx].atlas_rect.zw; - for (uint i = 0; i < scene_data.penumbra_shadow_samples; i++) { + for (uint i = 0; i < sc_penumbra_shadow_samples; i++) { vec2 suv = shadow_uv + (disk_rotation * scene_data.penumbra_shadow_kernel[i].xy) * uv_size; suv = clamp(suv, spot_lights.data[idx].atlas_rect.xy, clamp_max); float d = textureLod(sampler2D(shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), suv, 0.0).r; - if (d < z_norm) { + if (d < splane.z) { blocker_average += d; blocker_count += 1.0; } @@ -791,28 +783,23 @@ float light_process_spot_shadow(uint idx, vec3 vertex, vec3 normal) { uv_size *= penumbra; shadow = 0.0; - for (uint i = 0; i < scene_data.penumbra_shadow_samples; i++) { + for (uint i = 0; i < sc_penumbra_shadow_samples; i++) { vec2 suv = shadow_uv + (disk_rotation * scene_data.penumbra_shadow_kernel[i].xy) * uv_size; suv = clamp(suv, spot_lights.data[idx].atlas_rect.xy, clamp_max); - shadow += textureProj(sampler2DShadow(shadow_atlas, shadow_sampler), vec4(suv, z_norm, 1.0)); + shadow += textureProj(sampler2DShadow(shadow_atlas, shadow_sampler), vec4(suv, splane.z, 1.0)); } - shadow /= float(scene_data.penumbra_shadow_samples); + shadow /= float(sc_penumbra_shadow_samples); } else { //no blockers found, so no shadow shadow = 1.0; } - } else { -#endif //hard shadow - vec4 shadow_uv = vec4(splane.xy * spot_lights.data[idx].atlas_rect.zw + spot_lights.data[idx].atlas_rect.xy, splane.z, 1.0); - + vec3 shadow_uv = vec3(splane.xy * spot_lights.data[idx].atlas_rect.zw + spot_lights.data[idx].atlas_rect.xy, splane.z); shadow = sample_pcf_shadow(shadow_atlas, spot_lights.data[idx].soft_shadow_scale * scene_data.shadow_atlas_pixel_size, shadow_uv); -#ifdef USE_SOFT_SHADOWS } -#endif return shadow; } @@ -822,6 +809,18 @@ float light_process_spot_shadow(uint idx, vec3 vertex, vec3 normal) { return 1.0; } +vec2 normal_to_panorama(vec3 n) { + n = normalize(n); + vec2 panorama_coords = vec2(atan(n.x, n.z), acos(-n.y)); + + if (panorama_coords.x < 0.0) { + panorama_coords.x += M_PI * 2.0; + } + + panorama_coords /= vec2(M_PI * 2.0, M_PI); + return panorama_coords; +} + void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 vertex_ddx, vec3 vertex_ddy, vec3 f0, uint orms, float shadow, #ifdef LIGHT_BACKLIGHT_USED vec3 backlight, @@ -829,7 +828,6 @@ void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v #ifdef LIGHT_TRANSMITTANCE_USED vec4 transmittance_color, float transmittance_depth, - float transmittance_curve, float transmittance_boost, #endif #ifdef LIGHT_RIM_USED @@ -857,48 +855,66 @@ void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v vec3 color = spot_lights.data[idx].color; float specular_amount = spot_lights.data[idx].specular_amount; -#ifdef USE_SOFT_SHADOWS float size_A = 0.0; - if (spot_lights.data[idx].size > 0.0) { + if (sc_use_light_soft_shadows && spot_lights.data[idx].size > 0.0) { float t = spot_lights.data[idx].size / max(0.001, light_length); size_A = max(0.0, 1.0 - 1 / sqrt(1 + t * t)); } -#endif - - /* - if (spot_lights.data[idx].atlas_rect!=vec4(0.0)) { - //use projector texture - } - */ #ifdef LIGHT_TRANSMITTANCE_USED float transmittance_z = transmittance_depth; transmittance_color.a *= light_attenuation; { - splane = (spot_lights.data[idx].shadow_matrix * vec4(vertex - normalize(normal_interp) * spot_lights.data[idx].transmittance_bias, 1.0)); + vec4 splane = (spot_lights.data[idx].shadow_matrix * vec4(vertex - normalize(normal_interp) * spot_lights.data[idx].transmittance_bias, 1.0)); splane /= splane.w; splane.xy = splane.xy * spot_lights.data[idx].atlas_rect.zw + spot_lights.data[idx].atlas_rect.xy; float shadow_z = textureLod(sampler2D(shadow_atlas, material_samplers[SAMPLER_LINEAR_CLAMP]), splane.xy, 0.0).r; - //reconstruct depth - shadow_z /= spot_lights.data[idx].inv_radius; + + shadow_z = shadow_z * 2.0 - 1.0; + float z_far = 1.0 / spot_lights.data[idx].inv_radius; + float z_near = 0.01; + shadow_z = 2.0 * z_near * z_far / (z_far + z_near - shadow_z * (z_far - z_near)); + //distance to light plane float z = dot(spot_dir, -light_rel_vec); transmittance_z = z - shadow_z; } #endif //LIGHT_TRANSMITTANCE_USED + if (sc_use_light_projector && spot_lights.data[idx].projector_rect != vec4(0.0)) { + vec4 splane = (spot_lights.data[idx].shadow_matrix * vec4(vertex, 1.0)); + splane /= splane.w; + + vec2 proj_uv = normal_to_panorama(splane.xyz) * spot_lights.data[idx].projector_rect.zw; + + if (sc_projector_use_mipmaps) { + //ensure we have proper mipmaps + vec4 splane_ddx = (spot_lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddx, 1.0)); + splane_ddx /= splane_ddx.w; + vec2 proj_uv_ddx = normal_to_panorama(splane_ddx.xyz) * spot_lights.data[idx].projector_rect.zw - proj_uv; + + vec4 splane_ddy = (spot_lights.data[idx].shadow_matrix * vec4(vertex + vertex_ddy, 1.0)); + splane_ddy /= splane_ddy.w; + vec2 proj_uv_ddy = normal_to_panorama(splane_ddy.xyz) * spot_lights.data[idx].projector_rect.zw - proj_uv; + + vec4 proj = textureGrad(sampler2D(decal_atlas_srgb, light_projector_sampler), proj_uv + spot_lights.data[idx].projector_rect.xy, proj_uv_ddx, proj_uv_ddy); + color *= proj.rgb * proj.a; + } else { + vec4 proj = textureLod(sampler2D(decal_atlas_srgb, light_projector_sampler), proj_uv + spot_lights.data[idx].projector_rect.xy, 0.0); + color *= proj.rgb * proj.a; + } + } light_attenuation *= shadow; - light_compute(normal, normalize(light_rel_vec), eye_vec, color, light_attenuation, f0, orms, spot_lights.data[idx].specular_amount, + light_compute(normal, normalize(light_rel_vec), eye_vec, size_A, color, light_attenuation, f0, orms, spot_lights.data[idx].specular_amount, #ifdef LIGHT_BACKLIGHT_USED backlight, #endif #ifdef LIGHT_TRANSMITTANCE_USED transmittance_color, transmittance_depth, - transmittance_curve, transmittance_boost, transmittance_z, #endif @@ -911,9 +927,6 @@ void light_process_spot(uint idx, vec3 vertex, vec3 eye_vec, vec3 normal, vec3 v #ifdef LIGHT_ANISOTROPY_USED binormal, tangent, anisotropy, #endif -#ifdef USE_SOFT_SHADOW - size_A, -#endif #ifdef USE_SHADOW_TO_OPACITY alpha, #endif @@ -956,7 +969,7 @@ void reflection_process(uint ref_index, vec3 vertex, vec3 normal, float roughnes vec4 reflection; - reflection.rgb = textureLod(samplerCubeArray(reflection_atlas, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(local_ref_vec, reflections.data[ref_index].index), roughness * MAX_ROUGHNESS_LOD).rgb; + reflection.rgb = textureLod(samplerCubeArray(reflection_atlas, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), vec4(local_ref_vec, reflections.data[ref_index].index), roughness * MAX_ROUGHNESS_LOD).rgb * sc_luminance_multiplier; if (reflections.data[ref_index].exterior) { reflection.rgb = mix(specular_light, reflection.rgb, blend); diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_mobile.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_mobile.glsl index aa8a0b96c5..518b0a6c7f 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_mobile.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_mobile.glsl @@ -59,27 +59,27 @@ layout(location = 11) in vec4 weight_attrib; /* Varyings */ -layout(location = 0) out vec3 vertex_interp; +layout(location = 0) highp out vec3 vertex_interp; #ifdef NORMAL_USED -layout(location = 1) out vec3 normal_interp; +layout(location = 1) mediump out vec3 normal_interp; #endif #if defined(COLOR_USED) -layout(location = 2) out vec4 color_interp; +layout(location = 2) mediump out vec4 color_interp; #endif #ifdef UV_USED -layout(location = 3) out vec2 uv_interp; +layout(location = 3) mediump out vec2 uv_interp; #endif #if defined(UV2_USED) || defined(USE_LIGHTMAP) -layout(location = 4) out vec2 uv2_interp; +layout(location = 4) mediump out vec2 uv2_interp; #endif #if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) -layout(location = 5) out vec3 tangent_interp; -layout(location = 6) out vec3 binormal_interp; +layout(location = 5) mediump out vec3 tangent_interp; +layout(location = 6) mediump out vec3 binormal_interp; #endif #ifdef MATERIAL_UNIFORMS_USED @@ -92,7 +92,7 @@ layout(set = MATERIAL_UNIFORM_SET, binding = 0, std140) uniform MaterialUniforms #ifdef MODE_DUAL_PARABOLOID -layout(location = 8) out float dp_clip; +layout(location = 8) out highp float dp_clip; #endif @@ -124,7 +124,7 @@ void main() { mat3 world_normal_matrix; if (bool(draw_call.flags & INSTANCE_FLAGS_NON_UNIFORM_SCALE)) { - world_normal_matrix = inverse(mat3(world_matrix)); + world_normal_matrix = transpose(inverse(mat3(world_matrix))); } else { world_normal_matrix = mat3(world_matrix); } @@ -370,37 +370,70 @@ void main() { #VERSION_DEFINES +/* Specialization Constants */ + +#if !defined(MODE_RENDER_DEPTH) + +#if !defined(MODE_UNSHADED) + +layout(constant_id = 0) const bool sc_use_light_projector = false; +layout(constant_id = 1) const bool sc_use_light_soft_shadows = false; +layout(constant_id = 2) const bool sc_use_directional_soft_shadows = false; + +layout(constant_id = 3) const uint sc_soft_shadow_samples = 4; +layout(constant_id = 4) const uint sc_penumbra_shadow_samples = 4; + +layout(constant_id = 5) const uint sc_directional_soft_shadow_samples = 4; +layout(constant_id = 6) const uint sc_directional_penumbra_shadow_samples = 4; + +layout(constant_id = 8) const bool sc_projector_use_mipmaps = true; + +layout(constant_id = 9) const bool sc_disable_omni_lights = false; +layout(constant_id = 10) const bool sc_disable_spot_lights = false; +layout(constant_id = 11) const bool sc_disable_reflection_probes = false; +layout(constant_id = 12) const bool sc_disable_directional_lights = false; + +#endif //!MODE_UNSHADED + +layout(constant_id = 7) const bool sc_decal_use_mipmaps = true; +layout(constant_id = 13) const bool sc_disable_decals = false; +layout(constant_id = 14) const bool sc_disable_fog = false; + +#endif //!MODE_RENDER_DEPTH + +layout(constant_id = 15) const float sc_luminance_multiplier = 2.0; + /* Include our forward mobile UBOs definitions etc. */ #include "scene_forward_mobile_inc.glsl" /* Varyings */ -layout(location = 0) in vec3 vertex_interp; +layout(location = 0) highp in vec3 vertex_interp; #ifdef NORMAL_USED -layout(location = 1) in vec3 normal_interp; +layout(location = 1) mediump in vec3 normal_interp; #endif #if defined(COLOR_USED) -layout(location = 2) in vec4 color_interp; +layout(location = 2) mediump in vec4 color_interp; #endif #ifdef UV_USED -layout(location = 3) in vec2 uv_interp; +layout(location = 3) mediump in vec2 uv_interp; #endif #if defined(UV2_USED) || defined(USE_LIGHTMAP) -layout(location = 4) in vec2 uv2_interp; +layout(location = 4) mediump in vec2 uv2_interp; #endif #if defined(TANGENT_USED) || defined(NORMAL_MAP_USED) || defined(LIGHT_ANISOTROPY_USED) -layout(location = 5) in vec3 tangent_interp; -layout(location = 6) in vec3 binormal_interp; +layout(location = 5) mediump in vec3 tangent_interp; +layout(location = 6) mediump in vec3 binormal_interp; #endif #ifdef MODE_DUAL_PARABOLOID -layout(location = 8) in float dp_clip; +layout(location = 8) highp in float dp_clip; #endif @@ -462,7 +495,7 @@ layout(location = 0) out vec4 diffuse_buffer; //diffuse (rgb) and roughness layout(location = 1) out vec4 specular_buffer; //specular and SSS (subsurface scatter) #else -layout(location = 0) out vec4 frag_color; +layout(location = 0) out mediump vec4 frag_color; #endif // MODE_MULTIPLE_RENDER_TARGETS #endif // RENDER DEPTH @@ -471,6 +504,11 @@ layout(location = 0) out vec4 frag_color; #if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) +/* Make a default specular mode SPECULAR_SCHLICK_GGX. */ +#if !defined(SPECULAR_DISABLED) && !defined(SPECULAR_SCHLICK_GGX) && !defined(SPECULAR_BLINN) && !defined(SPECULAR_PHONG) && !defined(SPECULAR_TOON) +#define SPECULAR_SCHLICK_GGX +#endif + #include "scene_forward_lights_inc.glsl" #endif //!defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) @@ -543,7 +581,6 @@ void main() { vec3 backlight = vec3(0.0); vec4 transmittance_color = vec4(0.0); float transmittance_depth = 0.0; - float transmittance_curve = 1.0; float transmittance_boost = 0.0; float metallic = 0.0; float specular = 0.5; @@ -707,7 +744,7 @@ void main() { // to maximize VGPR usage // Draw "fixed" fog before volumetric fog to ensure volumetric fog can appear in front of the sky. - if (scene_data.fog_enabled) { + if (!sc_disable_fog && scene_data.fog_enabled) { fog = fog_process(vertex); } @@ -725,7 +762,7 @@ void main() { vec3 vertex_ddx = dFdx(vertex); vec3 vertex_ddy = dFdy(vertex); - { //Decals + if (!sc_disable_decals) { //Decals // must implement uint decal_indices = draw_call.decals.x; @@ -746,25 +783,35 @@ void main() { continue; //out of decal } - //we need ddx/ddy for mipmaps, so simulate them - vec2 ddx = (decals.data[decal_index].xform * vec4(vertex_ddx, 0.0)).xz; - vec2 ddy = (decals.data[decal_index].xform * vec4(vertex_ddy, 0.0)).xz; - float fade = pow(1.0 - (uv_local.y > 0.0 ? uv_local.y : -uv_local.y), uv_local.y > 0.0 ? decals.data[decal_index].upper_fade : decals.data[decal_index].lower_fade); if (decals.data[decal_index].normal_fade > 0.0) { fade *= smoothstep(decals.data[decal_index].normal_fade, 1.0, dot(normal_interp, decals.data[decal_index].normal) * 0.5 + 0.5); } + //we need ddx/ddy for mipmaps, so simulate them + vec2 ddx = (decals.data[decal_index].xform * vec4(vertex_ddx, 0.0)).xz; + vec2 ddy = (decals.data[decal_index].xform * vec4(vertex_ddy, 0.0)).xz; + if (decals.data[decal_index].albedo_rect != vec4(0.0)) { //has albedo - vec4 decal_albedo = textureGrad(sampler2D(decal_atlas_srgb, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uv_local.xz * decals.data[decal_index].albedo_rect.zw + decals.data[decal_index].albedo_rect.xy, ddx * decals.data[decal_index].albedo_rect.zw, ddy * decals.data[decal_index].albedo_rect.zw); + vec4 decal_albedo; + if (sc_decal_use_mipmaps) { + decal_albedo = textureGrad(sampler2D(decal_atlas_srgb, decal_sampler), uv_local.xz * decals.data[decal_index].albedo_rect.zw + decals.data[decal_index].albedo_rect.xy, ddx * decals.data[decal_index].albedo_rect.zw, ddy * decals.data[decal_index].albedo_rect.zw); + } else { + decal_albedo = textureLod(sampler2D(decal_atlas_srgb, decal_sampler), uv_local.xz * decals.data[decal_index].albedo_rect.zw + decals.data[decal_index].albedo_rect.xy, 0.0); + } decal_albedo *= decals.data[decal_index].modulate; decal_albedo.a *= fade; albedo = mix(albedo, decal_albedo.rgb, decal_albedo.a * decals.data[decal_index].albedo_mix); if (decals.data[decal_index].normal_rect != vec4(0.0)) { - vec3 decal_normal = textureGrad(sampler2D(decal_atlas, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uv_local.xz * decals.data[decal_index].normal_rect.zw + decals.data[decal_index].normal_rect.xy, ddx * decals.data[decal_index].normal_rect.zw, ddy * decals.data[decal_index].normal_rect.zw).xyz; + vec3 decal_normal; + if (sc_decal_use_mipmaps) { + decal_normal = textureGrad(sampler2D(decal_atlas, decal_sampler), uv_local.xz * decals.data[decal_index].normal_rect.zw + decals.data[decal_index].normal_rect.xy, ddx * decals.data[decal_index].normal_rect.zw, ddy * decals.data[decal_index].normal_rect.zw).xyz; + } else { + decal_normal = textureLod(sampler2D(decal_atlas, decal_sampler), uv_local.xz * decals.data[decal_index].normal_rect.zw + decals.data[decal_index].normal_rect.xy, 0.0).xyz; + } decal_normal.xy = decal_normal.xy * vec2(2.0, -2.0) - vec2(1.0, -1.0); //users prefer flipped y normal maps in most authoring software decal_normal.z = sqrt(max(0.0, 1.0 - dot(decal_normal.xy, decal_normal.xy))); //convert to view space, use xzy because y is up @@ -774,7 +821,12 @@ void main() { } if (decals.data[decal_index].orm_rect != vec4(0.0)) { - vec3 decal_orm = textureGrad(sampler2D(decal_atlas, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uv_local.xz * decals.data[decal_index].orm_rect.zw + decals.data[decal_index].orm_rect.xy, ddx * decals.data[decal_index].orm_rect.zw, ddy * decals.data[decal_index].orm_rect.zw).xyz; + vec3 decal_orm; + if (sc_decal_use_mipmaps) { + decal_orm = textureGrad(sampler2D(decal_atlas, decal_sampler), uv_local.xz * decals.data[decal_index].orm_rect.zw + decals.data[decal_index].orm_rect.xy, ddx * decals.data[decal_index].orm_rect.zw, ddy * decals.data[decal_index].orm_rect.zw).xyz; + } else { + decal_orm = textureLod(sampler2D(decal_atlas, decal_sampler), uv_local.xz * decals.data[decal_index].orm_rect.zw + decals.data[decal_index].orm_rect.xy, 0.0).xyz; + } ao = mix(ao, decal_orm.r, decal_albedo.a); roughness = mix(roughness, decal_orm.g, decal_albedo.a); metallic = mix(metallic, decal_orm.b, decal_albedo.a); @@ -783,7 +835,11 @@ void main() { if (decals.data[decal_index].emission_rect != vec4(0.0)) { //emission is additive, so its independent from albedo - emission += textureGrad(sampler2D(decal_atlas_srgb, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), uv_local.xz * decals.data[decal_index].emission_rect.zw + decals.data[decal_index].emission_rect.xy, ddx * decals.data[decal_index].emission_rect.zw, ddy * decals.data[decal_index].emission_rect.zw).xyz * decals.data[decal_index].emission_energy * fade; + if (sc_decal_use_mipmaps) { + emission += textureGrad(sampler2D(decal_atlas_srgb, decal_sampler), uv_local.xz * decals.data[decal_index].emission_rect.zw + decals.data[decal_index].emission_rect.xy, ddx * decals.data[decal_index].emission_rect.zw, ddy * decals.data[decal_index].emission_rect.zw).xyz * decals.data[decal_index].emission_energy * fade; + } else { + emission += textureLod(sampler2D(decal_atlas_srgb, decal_sampler), uv_local.xz * decals.data[decal_index].emission_rect.zw + decals.data[decal_index].emission_rect.xy, 0.0).xyz * decals.data[decal_index].emission_energy * fade; + } } } } //Decals @@ -793,7 +849,7 @@ void main() { #ifdef NORMAL_USED if (scene_data.roughness_limiter_enabled) { - //http://www.jp.square-enix.com/tech/library/pdf/ImprovedGeometricSpecularAA.pdf + //https://www.jp.square-enix.com/tech/library/pdf/ImprovedGeometricSpecularAA.pdf float roughness2 = roughness * roughness; vec3 dndu = dFdx(normal), dndv = dFdy(normal); float variance = scene_data.roughness_limiter_amount * (dot(dndu, dndu) + dot(dndv, dndv)); @@ -824,6 +880,8 @@ void main() { specular_light = textureLod(samplerCube(radiance_cubemap, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), ref_vec, roughness * MAX_ROUGHNESS_LOD).rgb; #endif //USE_RADIANCE_CUBEMAP_ARRAY + float horizon = min(1.0 + dot(ref_vec, normal), 1.0); + specular_light *= horizon * horizon; specular_light *= scene_data.ambient_light_color_energy.a; } @@ -921,7 +979,7 @@ void main() { // skipping ssao, do we remove ssao totally? - { //Reflection probes + if (!sc_disable_reflection_probes) { //Reflection probes vec4 reflection_accum = vec4(0.0, 0.0, 0.0, 0.0); vec4 ambient_accum = vec4(0.0, 0.0, 0.0, 0.0); @@ -962,7 +1020,7 @@ void main() { specular_light *= specular * metallic * albedo * 2.0; #else - // scales the specular reflections, needs to be be computed before lighting happens, + // scales the specular reflections, needs to be computed before lighting happens, // but after environment, GI, and reflection probes are added // Environment brdf approximation (Lazarov 2013) // see https://www.unrealengine.com/en-US/blog/physically-based-shading-on-mobile @@ -987,7 +1045,7 @@ void main() { // LIGHTING #if !defined(MODE_RENDER_DEPTH) && !defined(MODE_UNSHADED) - { //directional light + if (!sc_disable_directional_lights) { //directional light // Do shadow and lighting in two passes to reduce register pressure uint shadow0 = 0; @@ -1285,7 +1343,7 @@ void main() { blur_shadow(shadow); - light_compute(normal, directional_lights.data[i].direction, normalize(view), directional_lights.data[i].color * directional_lights.data[i].energy, shadow, f0, orms, 1.0, + light_compute(normal, directional_lights.data[i].direction, normalize(view), 0.0, directional_lights.data[i].color * directional_lights.data[i].energy, shadow, f0, orms, 1.0, #ifdef LIGHT_BACKLIGHT_USED backlight, #endif @@ -1293,7 +1351,6 @@ void main() { #ifdef LIGHT_TRANSMITTANCE_USED transmittance_color, transmittance_depth, - transmittance_curve, transmittance_boost, transmittance_z, #endif @@ -1318,7 +1375,7 @@ void main() { } } //directional light - { //omni lights + if (!sc_disable_omni_lights) { //omni lights uint light_indices = draw_call.omni_lights.x; for (uint i = 0; i < 8; i++) { uint light_index = light_indices & 0xFF; @@ -1332,7 +1389,7 @@ void main() { break; } - float shadow = light_process_omni_shadow(light_index, vertex, view); + float shadow = light_process_omni_shadow(light_index, vertex, normal); shadow = blur_shadow(shadow); @@ -1344,7 +1401,6 @@ void main() { #ifdef LIGHT_TRANSMITTANCE_USED transmittance_color, transmittance_depth, - transmittance_curve, transmittance_boost, #endif */ @@ -1366,7 +1422,7 @@ void main() { } } //omni lights - { //spot lights + if (!sc_disable_spot_lights) { //spot lights uint light_indices = draw_call.spot_lights.x; for (uint i = 0; i < 8; i++) { @@ -1381,7 +1437,7 @@ void main() { break; } - float shadow = light_process_spot_shadow(light_index, vertex, view); + float shadow = light_process_spot_shadow(light_index, vertex, normal); shadow = blur_shadow(shadow); @@ -1393,7 +1449,6 @@ void main() { #ifdef LIGHT_TRANSMITTANCE_USED transmittance_color, transmittance_depth, - transmittance_curve, transmittance_boost, #endif */ @@ -1498,12 +1553,15 @@ void main() { frag_color = vec4(albedo, alpha); #else // MODE_UNSHADED frag_color = vec4(emission + ambient_light + diffuse_light + specular_light, alpha); - //frag_color = vec4(1.0); #endif // MODE_UNSHADED // Draw "fixed" fog before volumetric fog to ensure volumetric fog can appear in front of the sky. frag_color.rgb = mix(frag_color.rgb, fog.rgb, fog.a); + // On mobile we use a UNORM buffer with 10bpp which results in a range from 0.0 - 1.0 resulting in HDR breaking + // We divide by sc_luminance_multiplier to support a range from 0.0 - 2.0 both increasing precision on bright and darker images + frag_color.rgb = frag_color.rgb / sc_luminance_multiplier; + #endif //MODE_MULTIPLE_RENDER_TARGETS #endif //MODE_RENDER_DEPTH diff --git a/servers/rendering/renderer_rd/shaders/scene_forward_mobile_inc.glsl b/servers/rendering/renderer_rd/shaders/scene_forward_mobile_inc.glsl index 7fcd84695d..dd8879acb4 100644 --- a/servers/rendering/renderer_rd/shaders/scene_forward_mobile_inc.glsl +++ b/servers/rendering/renderer_rd/shaders/scene_forward_mobile_inc.glsl @@ -16,12 +16,12 @@ /* don't exceed 128 bytes!! */ /* put instance data into our push content, not a array */ layout(push_constant, binding = 0, std430) uniform DrawCall { - mat4 transform; // 64 - 64 + highp mat4 transform; // 64 - 64 uint flags; // 04 - 68 uint instance_uniforms_ofs; //base offset in global buffer for instance variables // 04 - 72 uint gi_offset; //GI information when using lightmapping (VCT or lightmap index) // 04 - 76 uint layer_mask; // 04 - 80 - vec4 lightmap_uv_scale; // 16 - 96 doubles as uv_offset when needed + highp vec4 lightmap_uv_scale; // 16 - 96 doubles as uv_offset when needed uvec2 reflection_probes; // 08 - 104 uvec2 omni_lights; // 08 - 112 @@ -51,6 +51,10 @@ layout(set = 0, binding = 1) uniform sampler material_samplers[12]; layout(set = 0, binding = 2) uniform sampler shadow_sampler; +layout(set = 0, binding = 3) uniform sampler decal_sampler; +layout(set = 0, binding = 4) uniform sampler light_projector_sampler; + +#define INSTANCE_FLAGS_NON_UNIFORM_SCALE (1 << 5) #define INSTANCE_FLAGS_USE_GI_BUFFERS (1 << 6) #define INSTANCE_FLAGS_USE_SDFGI (1 << 7) #define INSTANCE_FLAGS_USE_LIGHTMAP_CAPTURE (1 << 8) @@ -65,24 +69,22 @@ layout(set = 0, binding = 2) uniform sampler shadow_sampler; //3 bits of stride #define INSTANCE_FLAGS_PARTICLE_TRAIL_MASK 0xFF -#define INSTANCE_FLAGS_NON_UNIFORM_SCALE (1 << 24) - -layout(set = 0, binding = 3, std430) restrict readonly buffer OmniLights { +layout(set = 0, binding = 5, std430) restrict readonly buffer OmniLights { LightData data[]; } omni_lights; -layout(set = 0, binding = 4, std430) restrict readonly buffer SpotLights { +layout(set = 0, binding = 6, std430) restrict readonly buffer SpotLights { LightData data[]; } spot_lights; -layout(set = 0, binding = 5, std430) restrict readonly buffer ReflectionProbeData { +layout(set = 0, binding = 7, std430) restrict readonly buffer ReflectionProbeData { ReflectionData data[]; } reflections; -layout(set = 0, binding = 6, std140) uniform DirectionalLights { +layout(set = 0, binding = 8, std140) uniform DirectionalLights { DirectionalLightData data[MAX_DIRECTIONAL_LIGHT_DATA_STRUCTS]; } directional_lights; @@ -91,103 +93,98 @@ directional_lights; #define LIGHTMAP_FLAG_USE_SPECULAR_DIRECTION 2 struct Lightmap { - mat3 normal_xform; + mediump mat3 normal_xform; }; -layout(set = 0, binding = 7, std140) restrict readonly buffer Lightmaps { +layout(set = 0, binding = 9, std140) restrict readonly buffer Lightmaps { Lightmap data[]; } lightmaps; struct LightmapCapture { - vec4 sh[9]; + mediump vec4 sh[9]; }; -layout(set = 0, binding = 8, std140) restrict readonly buffer LightmapCaptures { +layout(set = 0, binding = 10, std140) restrict readonly buffer LightmapCaptures { LightmapCapture data[]; } lightmap_captures; -layout(set = 0, binding = 9) uniform texture2D decal_atlas; -layout(set = 0, binding = 10) uniform texture2D decal_atlas_srgb; +layout(set = 0, binding = 11) uniform mediump texture2D decal_atlas; +layout(set = 0, binding = 12) uniform mediump texture2D decal_atlas_srgb; -layout(set = 0, binding = 11, std430) restrict readonly buffer Decals { +layout(set = 0, binding = 13, std430) restrict readonly buffer Decals { DecalData data[]; } decals; -layout(set = 0, binding = 12, std430) restrict readonly buffer GlobalVariableData { - vec4 data[]; +layout(set = 0, binding = 14, std430) restrict readonly buffer GlobalVariableData { + highp vec4 data[]; } global_variables; /* Set 1: Render Pass (changes per render pass) */ layout(set = 1, binding = 0, std140) uniform SceneData { - mat4 projection_matrix; - mat4 inv_projection_matrix; - mat4 camera_matrix; - mat4 inv_camera_matrix; + highp mat4 projection_matrix; + highp mat4 inv_projection_matrix; + highp mat4 camera_matrix; + highp mat4 inv_camera_matrix; // only used for multiview - mat4 projection_matrix_view[MAX_VIEWS]; - mat4 inv_projection_matrix_view[MAX_VIEWS]; - - vec2 viewport_size; - vec2 screen_pixel_size; + highp mat4 projection_matrix_view[MAX_VIEWS]; + highp mat4 inv_projection_matrix_view[MAX_VIEWS]; - //use vec4s because std140 doesnt play nice with vec2s, z and w are wasted - vec4 directional_penumbra_shadow_kernel[32]; - vec4 directional_soft_shadow_kernel[32]; - vec4 penumbra_shadow_kernel[32]; - vec4 soft_shadow_kernel[32]; + highp vec2 viewport_size; + highp vec2 screen_pixel_size; - uint directional_penumbra_shadow_samples; - uint directional_soft_shadow_samples; - uint penumbra_shadow_samples; - uint soft_shadow_samples; + // Use vec4s because std140 doesn't play nice with vec2s, z and w are wasted. + highp vec4 directional_penumbra_shadow_kernel[32]; + highp vec4 directional_soft_shadow_kernel[32]; + highp vec4 penumbra_shadow_kernel[32]; + highp vec4 soft_shadow_kernel[32]; - vec4 ambient_light_color_energy; + mediump vec4 ambient_light_color_energy; - float ambient_color_sky_mix; + mediump float ambient_color_sky_mix; bool use_ambient_light; bool use_ambient_cubemap; bool use_reflection_cubemap; - mat3 radiance_inverse_xform; + mediump mat3 radiance_inverse_xform; - vec2 shadow_atlas_pixel_size; - vec2 directional_shadow_pixel_size; + highp vec2 shadow_atlas_pixel_size; + highp vec2 directional_shadow_pixel_size; uint directional_light_count; - float dual_paraboloid_side; - float z_far; - float z_near; + mediump float dual_paraboloid_side; + highp float z_far; + highp float z_near; bool ssao_enabled; - float ssao_light_affect; - float ssao_ao_affect; + mediump float ssao_light_affect; + mediump float ssao_ao_affect; bool roughness_limiter_enabled; - float roughness_limiter_amount; - float roughness_limiter_limit; + mediump float roughness_limiter_amount; + mediump float roughness_limiter_limit; uvec2 roughness_limiter_pad; - vec4 ao_color; + mediump vec4 ao_color; bool fog_enabled; - float fog_density; - float fog_height; - float fog_height_density; + highp float fog_density; + highp float fog_height; + highp float fog_height_density; - vec3 fog_light_color; - float fog_sun_scatter; + mediump vec3 fog_light_color; + mediump float fog_sun_scatter; - float fog_aerial_perspective; + mediump float fog_aerial_perspective; bool material_uv2_mode; - float time; - float reflection_multiplier; // one normally, zero when rendering reflections + highp float time; + mediump float reflection_multiplier; // one normally, zero when rendering reflections bool pancake_shadows; uint pad1; @@ -198,30 +195,30 @@ scene_data; #ifdef USE_RADIANCE_CUBEMAP_ARRAY -layout(set = 1, binding = 2) uniform textureCubeArray radiance_cubemap; +layout(set = 1, binding = 2) uniform mediump textureCubeArray radiance_cubemap; #else -layout(set = 1, binding = 2) uniform textureCube radiance_cubemap; +layout(set = 1, binding = 2) uniform mediump textureCube radiance_cubemap; #endif -layout(set = 1, binding = 3) uniform textureCubeArray reflection_atlas; +layout(set = 1, binding = 3) uniform mediump textureCubeArray reflection_atlas; -layout(set = 1, binding = 4) uniform texture2D shadow_atlas; +layout(set = 1, binding = 4) uniform highp texture2D shadow_atlas; -layout(set = 1, binding = 5) uniform texture2D directional_shadow_atlas; +layout(set = 1, binding = 5) uniform highp texture2D directional_shadow_atlas; // this needs to change to providing just the lightmap we're using.. layout(set = 1, binding = 6) uniform texture2DArray lightmap_textures[MAX_LIGHTMAP_TEXTURES]; -layout(set = 1, binding = 9) uniform texture2D depth_buffer; -layout(set = 1, binding = 10) uniform texture2D color_buffer; +layout(set = 1, binding = 9) uniform highp texture2D depth_buffer; +layout(set = 1, binding = 10) uniform mediump texture2D color_buffer; /* Set 2 Skeleton & Instancing (can change per item) */ layout(set = 2, binding = 0, std430) restrict readonly buffer Transforms { - vec4 data[]; + highp vec4 data[]; } transforms; diff --git a/servers/rendering/renderer_rd/shaders/screen_space_reflection_scale.glsl b/servers/rendering/renderer_rd/shaders/screen_space_reflection_scale.glsl index 7e06516d90..2328effe7b 100644 --- a/servers/rendering/renderer_rd/shaders/screen_space_reflection_scale.glsl +++ b/servers/rendering/renderer_rd/shaders/screen_space_reflection_scale.glsl @@ -36,12 +36,12 @@ void main() { float divisor = 0.0; vec4 color; float depth; - vec3 normal; + vec4 normal; if (params.filtered) { color = vec4(0.0); depth = 0.0; - normal = vec3(0.0); + normal = vec4(0.0); for (int i = 0; i < 4; i++) { ivec2 ofs = ssC << 1; @@ -53,7 +53,9 @@ void main() { } color += texelFetch(source_ssr, ofs, 0); float d = texelFetch(source_depth, ofs, 0).r; - normal += texelFetch(source_normal, ofs, 0).xyz * 2.0 - 1.0; + vec4 nr = texelFetch(source_normal, ofs, 0); + normal.xyz += nr.xyz * 2.0 - 1.0; + normal.w += nr.w; d = d * 2.0 - 1.0; if (params.orthogonal) { @@ -66,11 +68,12 @@ void main() { color /= 4.0; depth /= 4.0; - normal = normalize(normal / 4.0) * 0.5 + 0.5; + normal.xyz = normalize(normal.xyz / 4.0) * 0.5 + 0.5; + normal.w /= 4.0; } else { color = texelFetch(source_ssr, ssC << 1, 0); depth = texelFetch(source_depth, ssC << 1, 0).r; - normal = texelFetch(source_normal, ssC << 1, 0).xyz; + normal = texelFetch(source_normal, ssC << 1, 0); depth = depth * 2.0 - 1.0; if (params.orthogonal) { @@ -83,5 +86,5 @@ void main() { imageStore(dest_ssr, ssC, color); imageStore(dest_depth, ssC, vec4(depth)); - imageStore(dest_normal, ssC, vec4(normal, 0.0)); + imageStore(dest_normal, ssC, normal); } diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_debug_probes.glsl b/servers/rendering/renderer_rd/shaders/sdfgi_debug_probes.glsl index 0eacbc5363..4290d5b869 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_debug_probes.glsl +++ b/servers/rendering/renderer_rd/shaders/sdfgi_debug_probes.glsl @@ -24,7 +24,7 @@ layout(push_constant, binding = 0, std430) uniform Params { } params; -// http://in4k.untergrund.net/html_articles/hugi_27_-_coding_corner_polaris_sphere_tessellation_101.htm +// https://in4k.untergrund.net/html_articles/hugi_27_-_coding_corner_polaris_sphere_tessellation_101.htm vec3 get_sphere_vertex(uint p_vertex_id) { float x_angle = float(p_vertex_id & 1u) + (p_vertex_id >> params.band_power); diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl b/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl index 99db35bb34..d6e5c6a92e 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl +++ b/servers/rendering/renderer_rd/shaders/sdfgi_direct_light.glsl @@ -20,10 +20,10 @@ layout(set = 0, binding = 3, std430) restrict readonly buffer DispatchData { dispatch_data; struct ProcessVoxel { - uint position; //xyz 7 bit packed, extra 11 bits for neigbours - uint albedo; //rgb bits 0-15 albedo, bits 16-21 are normal bits (set if geometry exists toward that side), extra 11 bits for neibhbours - uint light; //rgbe8985 encoded total saved light, extra 2 bits for neighbours - uint light_aniso; //55555 light anisotropy, extra 2 bits for neighbours + uint position; // xyz 7 bit packed, extra 11 bits for neighbors. + uint albedo; // rgb bits 0-15 albedo, bits 16-21 are normal bits (set if geometry exists toward that side), extra 11 bits for neighbors. + uint light; // rgbe8985 encoded total saved light, extra 2 bits for neighbors. + uint light_aniso; // 55555 light anisotropy, extra 2 bits for neighbors. //total neighbours: 26 }; diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl b/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl index bc376e9522..eedd28959c 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl +++ b/servers/rendering/renderer_rd/shaders/sdfgi_integrate.glsl @@ -266,9 +266,9 @@ void main() { } else if (params.sky_mode == SKY_MODE_SKY) { #ifdef USE_CUBEMAP_ARRAY - light.rgb = textureLod(samplerCubeArray(sky_irradiance, linear_sampler_mipmaps), vec4(ray_dir, 0.0), 2.0).rgb; //use second mipmap because we dont usually throw a lot of rays, so this compensates + light.rgb = textureLod(samplerCubeArray(sky_irradiance, linear_sampler_mipmaps), vec4(ray_dir, 0.0), 2.0).rgb; // Use second mipmap because we don't usually throw a lot of rays, so this compensates. #else - light.rgb = textureLod(samplerCube(sky_irradiance, linear_sampler_mipmaps), ray_dir, 2.0).rgb; //use second mipmap because we dont usually throw a lot of rays, so this compensates + light.rgb = textureLod(samplerCube(sky_irradiance, linear_sampler_mipmaps), ray_dir, 2.0).rgb; // Use second mipmap because we don't usually throw a lot of rays, so this compensates. #endif light.rgb *= params.sky_energy; light.a = 0.0; diff --git a/servers/rendering/renderer_rd/shaders/sdfgi_preprocess.glsl b/servers/rendering/renderer_rd/shaders/sdfgi_preprocess.glsl index aa4ded146f..4d9fa85a74 100644 --- a/servers/rendering/renderer_rd/shaders/sdfgi_preprocess.glsl +++ b/servers/rendering/renderer_rd/shaders/sdfgi_preprocess.glsl @@ -101,7 +101,7 @@ layout(set = 0, binding = 10, std430) restrict buffer DispatchData { dispatch_data; struct ProcessVoxel { - uint position; //xyz 7 bit packed, extra 11 bits for neigbours + uint position; // xyz 7 bit packed, extra 11 bits for neighbors. uint albedo; //rgb bits 0-15 albedo, bits 16-21 are normal bits (set if geometry exists toward that side), extra 11 bits for neibhbours uint light; //rgbe8985 encoded total saved light, extra 2 bits for neighbours uint light_aniso; //55555 light anisotropy, extra 2 bits for neighbours @@ -134,7 +134,7 @@ layout(set = 0, binding = 5, std430) restrict buffer readonly DispatchData { dispatch_data; struct ProcessVoxel { - uint position; //xyz 7 bit packed, extra 11 bits for neigbours + uint position; // xyz 7 bit packed, extra 11 bits for neighbors. uint albedo; //rgb bits 0-15 albedo, bits 16-21 are normal bits (set if geometry exists toward that side), extra 11 bits for neibhbours uint light; //rgbe8985 encoded total saved light, extra 2 bits for neighbours uint light_aniso; //55555 light anisotropy, extra 2 bits for neighbours @@ -183,7 +183,7 @@ void main() { ivec3 write_pos = read_pos + params.scroll; if (any(lessThan(write_pos, ivec3(0))) || any(greaterThanEqual(write_pos, ivec3(params.grid_size)))) { - return; //fits outside the 3D texture, dont do anything + return; // Fits outside the 3D texture, don't do anything. } uint albedo = ((src_process_voxels.data[index].albedo & 0x7FFF) << 1) | 1; //add solid bit diff --git a/servers/rendering/renderer_rd/shaders/sky.glsl b/servers/rendering/renderer_rd/shaders/sky.glsl index 41c6325bc5..d07a454ade 100644 --- a/servers/rendering/renderer_rd/shaders/sky.glsl +++ b/servers/rendering/renderer_rd/shaders/sky.glsl @@ -17,6 +17,8 @@ layout(push_constant, binding = 1, std430) uniform Params { vec4 projections[MAX_VIEWS]; vec4 position_multiplier; float time; + float luminance_multiplier; + float pad[2]; } params; @@ -55,6 +57,8 @@ layout(push_constant, binding = 1, std430) uniform Params { vec4 projections[MAX_VIEWS]; vec4 position_multiplier; float time; + float luminance_multiplier; + float pad[2]; } params; @@ -199,17 +203,17 @@ void main() { vec3 inverted_cube_normal = cube_normal; inverted_cube_normal.z *= -1.0; #ifdef USES_HALF_RES_COLOR - half_res_color = texture(samplerCube(half_res, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), inverted_cube_normal); + half_res_color = texture(samplerCube(half_res, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), inverted_cube_normal) * params.luminance_multiplier; #endif #ifdef USES_QUARTER_RES_COLOR - quarter_res_color = texture(samplerCube(quarter_res, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), inverted_cube_normal); + quarter_res_color = texture(samplerCube(quarter_res, material_samplers[SAMPLER_LINEAR_WITH_MIPMAPS_CLAMP]), inverted_cube_normal) * params.luminance_multiplier; #endif #else #ifdef USES_HALF_RES_COLOR - half_res_color = textureLod(sampler2D(half_res, material_samplers[SAMPLER_LINEAR_CLAMP]), uv, 0.0); + half_res_color = textureLod(sampler2D(half_res, material_samplers[SAMPLER_LINEAR_CLAMP]), uv, 0.0) * params.luminance_multiplier; #endif #ifdef USES_QUARTER_RES_COLOR - quarter_res_color = textureLod(sampler2D(quarter_res, material_samplers[SAMPLER_LINEAR_CLAMP]), uv, 0.0); + quarter_res_color = textureLod(sampler2D(quarter_res, material_samplers[SAMPLER_LINEAR_CLAMP]), uv, 0.0) * params.luminance_multiplier; #endif #endif @@ -246,4 +250,7 @@ void main() { if (!AT_CUBEMAP_PASS && !AT_HALF_RES_PASS && !AT_QUARTER_RES_PASS) { frag_color.a = 0.0; } + + // For mobile renderer we're dividing by 2.0 as we're using a UNORM buffer + frag_color.rgb = frag_color.rgb / params.luminance_multiplier; } diff --git a/servers/rendering/renderer_rd/shaders/tonemap.glsl b/servers/rendering/renderer_rd/shaders/tonemap.glsl index 23f83b3b9c..4411587116 100644 --- a/servers/rendering/renderer_rd/shaders/tonemap.glsl +++ b/servers/rendering/renderer_rd/shaders/tonemap.glsl @@ -35,13 +35,17 @@ void main() { layout(location = 0) in vec2 uv_interp; -#ifdef MULTIVIEW +#ifdef SUBPASS +layout(input_attachment_index = 0, set = 0, binding = 0) uniform subpassInput input_color; +#elif defined(MULTIVIEW) layout(set = 0, binding = 0) uniform sampler2DArray source_color; #else layout(set = 0, binding = 0) uniform sampler2D source_color; #endif + layout(set = 1, binding = 0) uniform sampler2D source_auto_exposure; layout(set = 2, binding = 0) uniform sampler2D source_glow; + #ifdef USE_1D_LUT layout(set = 3, binding = 0) uniform sampler2D source_color_correction; #else @@ -67,7 +71,7 @@ layout(push_constant, binding = 1, std430) uniform Params { float exposure; float white; float auto_exposure_grey; - uint pad2; + float luminance_multiplier; vec2 pixel_size; bool use_fxaa; @@ -180,10 +184,6 @@ vec3 tonemap_aces(vec3 color, float white) { } vec3 tonemap_reinhard(vec3 color, float white) { - // Ensure color values are positive. - // They can be negative in the case of negative lights, which leads to undesired behavior. - color = max(vec3(0.0), color); - return (white * color + color) / (color * white + white); } @@ -207,7 +207,7 @@ vec3 apply_tonemapping(vec3 color, float white) { // inputs are LINEAR, always o return tonemap_reinhard(color, white); } else if (params.tonemapper == TONEMAPPER_FILMIC) { return tonemap_filmic(color, white); - } else { //aces + } else { // TONEMAPPER_ACES return tonemap_aces(color, white); } } @@ -291,21 +291,22 @@ vec3 apply_color_correction(vec3 color) { } #endif +#ifndef SUBPASS vec3 do_fxaa(vec3 color, float exposure, vec2 uv_interp) { const float FXAA_REDUCE_MIN = (1.0 / 128.0); const float FXAA_REDUCE_MUL = (1.0 / 8.0); const float FXAA_SPAN_MAX = 8.0; #ifdef MULTIVIEW - vec3 rgbNW = textureLod(source_color, vec3(uv_interp + vec2(-1.0, -1.0) * params.pixel_size, ViewIndex), 0.0).xyz * exposure; - vec3 rgbNE = textureLod(source_color, vec3(uv_interp + vec2(1.0, -1.0) * params.pixel_size, ViewIndex), 0.0).xyz * exposure; - vec3 rgbSW = textureLod(source_color, vec3(uv_interp + vec2(-1.0, 1.0) * params.pixel_size, ViewIndex), 0.0).xyz * exposure; - vec3 rgbSE = textureLod(source_color, vec3(uv_interp + vec2(1.0, 1.0) * params.pixel_size, ViewIndex), 0.0).xyz * exposure; + vec3 rgbNW = textureLod(source_color, vec3(uv_interp + vec2(-1.0, -1.0) * params.pixel_size, ViewIndex), 0.0).xyz * exposure * params.luminance_multiplier; + vec3 rgbNE = textureLod(source_color, vec3(uv_interp + vec2(1.0, -1.0) * params.pixel_size, ViewIndex), 0.0).xyz * exposure * params.luminance_multiplier; + vec3 rgbSW = textureLod(source_color, vec3(uv_interp + vec2(-1.0, 1.0) * params.pixel_size, ViewIndex), 0.0).xyz * exposure * params.luminance_multiplier; + vec3 rgbSE = textureLod(source_color, vec3(uv_interp + vec2(1.0, 1.0) * params.pixel_size, ViewIndex), 0.0).xyz * exposure * params.luminance_multiplier; #else - vec3 rgbNW = textureLod(source_color, uv_interp + vec2(-1.0, -1.0) * params.pixel_size, 0.0).xyz * exposure; - vec3 rgbNE = textureLod(source_color, uv_interp + vec2(1.0, -1.0) * params.pixel_size, 0.0).xyz * exposure; - vec3 rgbSW = textureLod(source_color, uv_interp + vec2(-1.0, 1.0) * params.pixel_size, 0.0).xyz * exposure; - vec3 rgbSE = textureLod(source_color, uv_interp + vec2(1.0, 1.0) * params.pixel_size, 0.0).xyz * exposure; + vec3 rgbNW = textureLod(source_color, uv_interp + vec2(-1.0, -1.0) * params.pixel_size, 0.0).xyz * exposure * params.luminance_multiplier; + vec3 rgbNE = textureLod(source_color, uv_interp + vec2(1.0, -1.0) * params.pixel_size, 0.0).xyz * exposure * params.luminance_multiplier; + vec3 rgbSW = textureLod(source_color, uv_interp + vec2(-1.0, 1.0) * params.pixel_size, 0.0).xyz * exposure * params.luminance_multiplier; + vec3 rgbSE = textureLod(source_color, uv_interp + vec2(1.0, 1.0) * params.pixel_size, 0.0).xyz * exposure * params.luminance_multiplier; #endif vec3 rgbM = color; vec3 luma = vec3(0.299, 0.587, 0.114); @@ -332,11 +333,11 @@ vec3 do_fxaa(vec3 color, float exposure, vec2 uv_interp) { params.pixel_size; #ifdef MULTIVIEW - vec3 rgbA = 0.5 * exposure * (textureLod(source_color, vec3(uv_interp + dir * (1.0 / 3.0 - 0.5), ViewIndex), 0.0).xyz + textureLod(source_color, vec3(uv_interp + dir * (2.0 / 3.0 - 0.5), ViewIndex), 0.0).xyz); - vec3 rgbB = rgbA * 0.5 + 0.25 * exposure * (textureLod(source_color, vec3(uv_interp + dir * -0.5, ViewIndex), 0.0).xyz + textureLod(source_color, vec3(uv_interp + dir * 0.5, ViewIndex), 0.0).xyz); + vec3 rgbA = 0.5 * exposure * (textureLod(source_color, vec3(uv_interp + dir * (1.0 / 3.0 - 0.5), ViewIndex), 0.0).xyz + textureLod(source_color, vec3(uv_interp + dir * (2.0 / 3.0 - 0.5), ViewIndex), 0.0).xyz) * params.luminance_multiplier; + vec3 rgbB = rgbA * 0.5 + 0.25 * exposure * (textureLod(source_color, vec3(uv_interp + dir * -0.5, ViewIndex), 0.0).xyz + textureLod(source_color, vec3(uv_interp + dir * 0.5, ViewIndex), 0.0).xyz) * params.luminance_multiplier; #else - vec3 rgbA = 0.5 * exposure * (textureLod(source_color, uv_interp + dir * (1.0 / 3.0 - 0.5), 0.0).xyz + textureLod(source_color, uv_interp + dir * (2.0 / 3.0 - 0.5), 0.0).xyz); - vec3 rgbB = rgbA * 0.5 + 0.25 * exposure * (textureLod(source_color, uv_interp + dir * -0.5, 0.0).xyz + textureLod(source_color, uv_interp + dir * 0.5, 0.0).xyz); + vec3 rgbA = 0.5 * exposure * (textureLod(source_color, uv_interp + dir * (1.0 / 3.0 - 0.5), 0.0).xyz + textureLod(source_color, uv_interp + dir * (2.0 / 3.0 - 0.5), 0.0).xyz) * params.luminance_multiplier; + vec3 rgbB = rgbA * 0.5 + 0.25 * exposure * (textureLod(source_color, uv_interp + dir * -0.5, 0.0).xyz + textureLod(source_color, uv_interp + dir * 0.5, 0.0).xyz) * params.luminance_multiplier; #endif float lumaB = dot(rgbB, luma); @@ -346,8 +347,9 @@ vec3 do_fxaa(vec3 color, float exposure, vec2 uv_interp) { return rgbB; } } +#endif // !SUBPASS -// From http://alex.vlachos.com/graphics/Alex_Vlachos_Advanced_VR_Rendering_GDC2015.pdf +// From https://alex.vlachos.com/graphics/Alex_Vlachos_Advanced_VR_Rendering_GDC2015.pdf // and https://www.shadertoy.com/view/MslGR8 (5th one starting from the bottom) // NOTE: `frag_coord` is in pixels (i.e. not normalized UV). vec3 screen_space_dither(vec2 frag_coord) { @@ -360,45 +362,56 @@ vec3 screen_space_dither(vec2 frag_coord) { } void main() { -#ifdef MULTIVIEW - vec3 color = textureLod(source_color, vec3(uv_interp, ViewIndex), 0.0f).rgb; +#ifdef SUBPASS + // SUBPASS and MULTIVIEW can be combined but in that case we're already reading from the correct layer + vec3 color = subpassLoad(input_color).rgb * params.luminance_multiplier; +#elif defined(MULTIVIEW) + vec3 color = textureLod(source_color, vec3(uv_interp, ViewIndex), 0.0f).rgb * params.luminance_multiplier; #else - vec3 color = textureLod(source_color, uv_interp, 0.0f).rgb; + vec3 color = textureLod(source_color, uv_interp, 0.0f).rgb * params.luminance_multiplier; #endif // Exposure float exposure = params.exposure; +#ifndef SUBPASS if (params.use_auto_exposure) { - exposure *= 1.0 / (texelFetch(source_auto_exposure, ivec2(0, 0), 0).r / params.auto_exposure_grey); + exposure *= 1.0 / (texelFetch(source_auto_exposure, ivec2(0, 0), 0).r * params.luminance_multiplier / params.auto_exposure_grey); } +#endif color *= exposure; // Early Tonemap & SRGB Conversion - +#ifndef SUBPASS if (params.use_glow && params.glow_mode == GLOW_MODE_MIX) { - vec3 glow = gather_glow(source_glow, uv_interp); + vec3 glow = gather_glow(source_glow, uv_interp) * params.luminance_multiplier; color.rgb = mix(color.rgb, glow, params.glow_intensity); } if (params.use_fxaa) { color = do_fxaa(color, exposure, uv_interp); } +#endif + if (params.use_debanding) { // For best results, debanding should be done before tonemapping. // Otherwise, we're adding noise to an already-quantized image. color += screen_space_dither(gl_FragCoord.xy); } - color = apply_tonemapping(color, params.white); + + // Ensure color values passed to tonemappers are positive. + // They can be negative in the case of negative lights, which leads to undesired behavior. + color = apply_tonemapping(max(vec3(0.0), color), params.white); color = linear_to_srgb(color); // regular linear -> SRGB conversion +#ifndef SUBPASS // Glow if (params.use_glow && params.glow_mode != GLOW_MODE_MIX) { - vec3 glow = gather_glow(source_glow, uv_interp) * params.glow_intensity; + vec3 glow = gather_glow(source_glow, uv_interp) * params.glow_intensity * params.luminance_multiplier; // high dynamic range -> SRGB glow = apply_tonemapping(glow, params.white); @@ -406,6 +419,7 @@ void main() { color = apply_glow(color, glow); } +#endif // Additional effects diff --git a/servers/rendering/renderer_rd/shaders/voxel_gi.glsl b/servers/rendering/renderer_rd/shaders/voxel_gi.glsl index 49a493cdc7..779f04ed35 100644 --- a/servers/rendering/renderer_rd/shaders/voxel_gi.glsl +++ b/servers/rendering/renderer_rd/shaders/voxel_gi.glsl @@ -71,11 +71,6 @@ lights; layout(set = 0, binding = 5) uniform texture3D color_texture; -#ifdef MODE_ANISOTROPIC -layout(set = 0, binding = 7) uniform texture3D aniso_pos_texture; -layout(set = 0, binding = 8) uniform texture3D aniso_neg_texture; -#endif // MODE ANISOTROPIC - #endif // MODE_SECOND_BOUNCE #ifndef MODE_DYNAMIC @@ -110,13 +105,6 @@ layout(set = 0, binding = 10) uniform sampler texture_sampler; layout(rgba8, set = 0, binding = 5) uniform restrict writeonly image3D color_tex; -#ifdef MODE_ANISOTROPIC - -layout(r16ui, set = 0, binding = 6) uniform restrict writeonly uimage3D aniso_pos_tex; -layout(r16ui, set = 0, binding = 7) uniform restrict writeonly uimage3D aniso_neg_tex; - -#endif - #endif #ifdef MODE_DYNAMIC @@ -170,13 +158,6 @@ layout(r32f, set = 0, binding = 8) uniform restrict writeonly image2D depth; layout(rgba8, set = 0, binding = 11) uniform restrict image3D color_texture; -#ifdef MODE_ANISOTROPIC - -layout(r16ui, set = 0, binding = 12) uniform restrict writeonly uimage3D aniso_pos_texture; -layout(r16ui, set = 0, binding = 13) uniform restrict writeonly uimage3D aniso_neg_texture; - -#endif // MODE ANISOTROPIC - #endif //MODE_DYNAMIC_SHRINK_PLOT #endif // MODE_DYNAMIC_SHRINK @@ -374,12 +355,7 @@ void main() { vec3 emission = vec3(uvec3(cell_data.data[cell_index].emission & 0x1ff, (cell_data.data[cell_index].emission >> 9) & 0x1ff, (cell_data.data[cell_index].emission >> 18) & 0x1ff)) * pow(2.0, float(cell_data.data[cell_index].emission >> 27) - 15.0 - 9.0); vec3 normal = unpackSnorm4x8(cell_data.data[cell_index].normal).xyz; -#ifdef MODE_ANISOTROPIC - vec3 accum[6] = vec3[](vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0)); - const vec3 accum_dirs[6] = vec3[](vec3(1.0, 0.0, 0.0), vec3(-1.0, 0.0, 0.0), vec3(0.0, 1.0, 0.0), vec3(0.0, -1.0, 0.0), vec3(0.0, 0.0, 1.0), vec3(0.0, 0.0, -1.0)); -#else vec3 accum = vec3(0.0); -#endif for (uint i = 0; i < params.light_count; i++) { vec3 light; @@ -390,38 +366,16 @@ void main() { light *= albedo.rgb; -#ifdef MODE_ANISOTROPIC - for (uint j = 0; j < 6; j++) { - accum[j] += max(0.0, dot(accum_dirs[j], -light_dir)) * light; - } -#else if (length(normal) > 0.2) { accum += max(0.0, dot(normal, -light_dir)) * light; } else { //all directions accum += light; } -#endif } -#ifdef MODE_ANISOTROPIC - - for (uint i = 0; i < 6; i++) { - vec3 light = accum[i]; - if (length(normal) > 0.2) { - light += max(0.0, dot(accum_dirs[i], -normal)) * emission; - } else { - light += emission; - } - - outputs.data[cell_index * 6 + i] = vec4(light, 0.0); - } - -#else outputs.data[cell_index] = vec4(accum + emission, 0.0); -#endif - #endif //MODE_COMPUTE_LIGHT /////////////////SECOND BOUNCE/////////////////////////////// @@ -431,32 +385,8 @@ void main() { ivec3 ipos = ivec3(posu); vec4 normal = unpackSnorm4x8(cell_data.data[cell_index].normal); -#ifdef MODE_ANISOTROPIC - vec3 accum[6]; - const vec3 accum_dirs[6] = vec3[](vec3(1.0, 0.0, 0.0), vec3(-1.0, 0.0, 0.0), vec3(0.0, 1.0, 0.0), vec3(0.0, -1.0, 0.0), vec3(0.0, 0.0, 1.0), vec3(0.0, 0.0, -1.0)); - - /*vec3 src_color = texelFetch(sampler3D(color_texture,texture_sampler),ipos,0).rgb * params.dynamic_range; - vec3 src_aniso_pos = texelFetch(sampler3D(aniso_pos_texture,texture_sampler),ipos,0).rgb; - vec3 src_anisp_neg = texelFetch(sampler3D(anisp_neg_texture,texture_sampler),ipos,0).rgb; - accum[0]=src_col * src_aniso_pos.x; - accum[1]=src_col * src_aniso_neg.x; - accum[2]=src_col * src_aniso_pos.y; - accum[3]=src_col * src_aniso_neg.y; - accum[4]=src_col * src_aniso_pos.z; - accum[5]=src_col * src_aniso_neg.z;*/ - - accum[0] = outputs.data[cell_index * 6 + 0].rgb; - accum[1] = outputs.data[cell_index * 6 + 1].rgb; - accum[2] = outputs.data[cell_index * 6 + 2].rgb; - accum[3] = outputs.data[cell_index * 6 + 3].rgb; - accum[4] = outputs.data[cell_index * 6 + 4].rgb; - accum[5] = outputs.data[cell_index * 6 + 5].rgb; - -#else vec3 accum = outputs.data[cell_index].rgb; -#endif - if (length(normal.xyz) > 0.2) { vec3 v0 = abs(normal.z) < 0.999 ? vec3(0.0, 0.0, 1.0) : vec3(0.0, 1.0, 0.0); vec3 tangent = normalize(cross(v0, normal.xyz)); @@ -484,9 +414,6 @@ void main() { float max_distance = length(vec3(params.limits)); vec3 cell_size = 1.0 / vec3(params.limits); -#ifdef MODE_ANISOTROPIC - vec3 aniso_normal = mix(direction, normal.xyz, params.aniso_strength); -#endif while (dist < max_distance && color.a < 0.95) { float diameter = max(1.0, 2.0 * tan_half_angle * dist); vec3 uvw_pos = (pos + dist * direction) * cell_size; @@ -498,42 +425,18 @@ void main() { float log2_diameter = log2(diameter); vec4 scolor = textureLod(sampler3D(color_texture, texture_sampler), uvw_pos, log2_diameter); -#ifdef MODE_ANISOTROPIC - - vec3 aniso_neg = textureLod(sampler3D(aniso_neg_texture, texture_sampler), uvw_pos, log2_diameter).rgb; - vec3 aniso_pos = textureLod(sampler3D(aniso_pos_texture, texture_sampler), uvw_pos, log2_diameter).rgb; - - scolor.rgb *= dot(max(vec3(0.0), (aniso_normal * aniso_pos)), vec3(1.0)) + dot(max(vec3(0.0), (-aniso_normal * aniso_neg)), vec3(1.0)); -#endif float a = (1.0 - color.a); color += a * scolor; dist += half_diameter; } } color *= cone_weights[i] * vec4(albedo.rgb, 1.0) * params.dynamic_range; //restore range -#ifdef MODE_ANISOTROPIC - for (uint j = 0; j < 6; j++) { - accum[j] += max(0.0, dot(accum_dirs[j], direction)) * color.rgb; - } -#else accum += color.rgb; -#endif } } -#ifdef MODE_ANISOTROPIC - - outputs.data[cell_index * 6 + 0] = vec4(accum[0], 0.0); - outputs.data[cell_index * 6 + 1] = vec4(accum[1], 0.0); - outputs.data[cell_index * 6 + 2] = vec4(accum[2], 0.0); - outputs.data[cell_index * 6 + 3] = vec4(accum[3], 0.0); - outputs.data[cell_index * 6 + 4] = vec4(accum[4], 0.0); - outputs.data[cell_index * 6 + 5] = vec4(accum[5], 0.0); -#else outputs.data[cell_index] = vec4(accum, 0.0); -#endif - #endif // MODE_SECOND_BOUNCE /////////////////UPDATE MIPMAPS/////////////////////////////// @@ -541,45 +444,20 @@ void main() { #ifdef MODE_UPDATE_MIPMAPS { -#ifdef MODE_ANISOTROPIC - vec3 light_accum[6] = vec3[](vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0), vec3(0.0)); -#else vec3 light_accum = vec3(0.0); -#endif float count = 0.0; for (uint i = 0; i < 8; i++) { uint child_index = cell_children.data[cell_index].children[i]; if (child_index == NO_CHILDREN) { continue; } -#ifdef MODE_ANISOTROPIC - light_accum[0] += outputs.data[child_index * 6 + 0].rgb; - light_accum[1] += outputs.data[child_index * 6 + 1].rgb; - light_accum[2] += outputs.data[child_index * 6 + 2].rgb; - light_accum[3] += outputs.data[child_index * 6 + 3].rgb; - light_accum[4] += outputs.data[child_index * 6 + 4].rgb; - light_accum[5] += outputs.data[child_index * 6 + 5].rgb; - -#else light_accum += outputs.data[child_index].rgb; -#endif - count += 1.0; } float divisor = mix(8.0, count, params.propagation); -#ifdef MODE_ANISOTROPIC - outputs.data[cell_index * 6 + 0] = vec4(light_accum[0] / divisor, 0.0); - outputs.data[cell_index * 6 + 1] = vec4(light_accum[1] / divisor, 0.0); - outputs.data[cell_index * 6 + 2] = vec4(light_accum[2] / divisor, 0.0); - outputs.data[cell_index * 6 + 3] = vec4(light_accum[3] / divisor, 0.0); - outputs.data[cell_index * 6 + 4] = vec4(light_accum[4] / divisor, 0.0); - outputs.data[cell_index * 6 + 5] = vec4(light_accum[5] / divisor, 0.0); - -#else outputs.data[cell_index] = vec4(light_accum / divisor, 0.0); -#endif } #endif @@ -587,40 +465,7 @@ void main() { #ifdef MODE_WRITE_TEXTURE { -#ifdef MODE_ANISOTROPIC - vec3 accum_total = vec3(0.0); - accum_total += outputs.data[cell_index * 6 + 0].rgb; - accum_total += outputs.data[cell_index * 6 + 1].rgb; - accum_total += outputs.data[cell_index * 6 + 2].rgb; - accum_total += outputs.data[cell_index * 6 + 3].rgb; - accum_total += outputs.data[cell_index * 6 + 4].rgb; - accum_total += outputs.data[cell_index * 6 + 5].rgb; - - float accum_total_energy = max(dot(accum_total, GREY_VEC), 0.00001); - vec3 iso_positive = vec3(dot(outputs.data[cell_index * 6 + 0].rgb, GREY_VEC), dot(outputs.data[cell_index * 6 + 2].rgb, GREY_VEC), dot(outputs.data[cell_index * 6 + 4].rgb, GREY_VEC)) / vec3(accum_total_energy); - vec3 iso_negative = vec3(dot(outputs.data[cell_index * 6 + 1].rgb, GREY_VEC), dot(outputs.data[cell_index * 6 + 3].rgb, GREY_VEC), dot(outputs.data[cell_index * 6 + 5].rgb, GREY_VEC)) / vec3(accum_total_energy); - - { - uint aniso_pos = uint(clamp(iso_positive.b * 31.0, 0.0, 31.0)); - aniso_pos |= uint(clamp(iso_positive.g * 63.0, 0.0, 63.0)) << 5; - aniso_pos |= uint(clamp(iso_positive.r * 31.0, 0.0, 31.0)) << 11; - imageStore(aniso_pos_tex, ivec3(posu), uvec4(aniso_pos)); - } - - { - uint aniso_neg = uint(clamp(iso_negative.b * 31.0, 0.0, 31.0)); - aniso_neg |= uint(clamp(iso_negative.g * 63.0, 0.0, 63.0)) << 5; - aniso_neg |= uint(clamp(iso_negative.r * 31.0, 0.0, 31.0)) << 11; - imageStore(aniso_neg_tex, ivec3(posu), uvec4(aniso_neg)); - } - - imageStore(color_tex, ivec3(posu), vec4(accum_total / params.dynamic_range, albedo.a)); - -#else - imageStore(color_tex, ivec3(posu), vec4(outputs.data[cell_index].rgb / params.dynamic_range, albedo.a)); - -#endif } #endif @@ -763,13 +608,6 @@ void main() { color.rgb /= params.dynamic_range; imageStore(color_texture, pos3d, color); //imageStore(color_texture,pos3d,vec4(1,1,1,1)); - -#ifdef MODE_ANISOTROPIC - //do not care about anisotropy for dynamic objects, just store full lit in all directions - imageStore(aniso_pos_texture, pos3d, uvec4(0xFFFF)); - imageStore(aniso_neg_texture, pos3d, uvec4(0xFFFF)); - -#endif // ANISOTROPIC } #endif // MODE_DYNAMIC_SHRINK_PLOT } diff --git a/servers/rendering/renderer_rd/shaders/voxel_gi_debug.glsl b/servers/rendering/renderer_rd/shaders/voxel_gi_debug.glsl index 7d4d72967a..281c496df3 100644 --- a/servers/rendering/renderer_rd/shaders/voxel_gi_debug.glsl +++ b/servers/rendering/renderer_rd/shaders/voxel_gi_debug.glsl @@ -20,11 +20,6 @@ layout(set = 0, binding = 2) uniform texture3D color_tex; layout(set = 0, binding = 3) uniform sampler tex_sampler; -#ifdef USE_ANISOTROPY -layout(set = 0, binding = 4) uniform texture3D aniso_pos_tex; -layout(set = 0, binding = 5) uniform texture3D aniso_neg_tex; -#endif - layout(push_constant, binding = 0, std430) uniform Params { mat4 projection; uint cell_offset; |